{ "best_metric": null, "best_model_checkpoint": null, "epoch": 1.0, "global_step": 10144, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 6.557377049180328e-06, "loss": 9.7891, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.3114754098360657e-05, "loss": 9.8047, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.9672131147540985e-05, "loss": 9.7422, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.6229508196721314e-05, "loss": 9.7891, "step": 4 }, { "epoch": 0.0, "learning_rate": 3.278688524590164e-05, "loss": 9.75, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.934426229508197e-05, "loss": 9.6406, "step": 6 }, { "epoch": 0.0, "learning_rate": 4.590163934426229e-05, "loss": 9.7812, "step": 7 }, { "epoch": 0.0, "learning_rate": 5.245901639344263e-05, "loss": 9.7109, "step": 8 }, { "epoch": 0.0, "learning_rate": 5.9016393442622956e-05, "loss": 9.4766, "step": 9 }, { "epoch": 0.0, "learning_rate": 6.557377049180328e-05, "loss": 9.5469, "step": 10 }, { "epoch": 0.0, "learning_rate": 7.213114754098361e-05, "loss": 9.5234, "step": 11 }, { "epoch": 0.0, "learning_rate": 7.868852459016394e-05, "loss": 9.3359, "step": 12 }, { "epoch": 0.0, "learning_rate": 8.524590163934426e-05, "loss": 9.3594, "step": 13 }, { "epoch": 0.0, "learning_rate": 9.180327868852458e-05, "loss": 9.2109, "step": 14 }, { "epoch": 0.0, "learning_rate": 9.836065573770491e-05, "loss": 9.2031, "step": 15 }, { "epoch": 0.0, "learning_rate": 0.00010491803278688525, "loss": 9.0078, "step": 16 }, { "epoch": 0.0, "learning_rate": 0.00011147540983606558, "loss": 8.8203, "step": 17 }, { "epoch": 0.0, "learning_rate": 0.00011803278688524591, "loss": 8.8281, "step": 18 }, { "epoch": 0.0, "learning_rate": 0.00012459016393442623, "loss": 8.7891, "step": 19 }, { "epoch": 0.0, "learning_rate": 0.00013114754098360657, "loss": 8.6094, "step": 20 }, { "epoch": 0.0, "learning_rate": 0.0001377049180327869, "loss": 8.6016, "step": 21 }, { "epoch": 0.0, "learning_rate": 0.00014426229508196722, "loss": 8.5078, "step": 22 }, { "epoch": 0.0, "learning_rate": 0.00015081967213114757, "loss": 8.4375, "step": 23 }, { "epoch": 0.0, "learning_rate": 0.00015737704918032788, "loss": 8.25, "step": 24 }, { "epoch": 0.0, "learning_rate": 0.0001639344262295082, "loss": 8.1172, "step": 25 }, { "epoch": 0.0, "learning_rate": 0.0001704918032786885, "loss": 7.9492, "step": 26 }, { "epoch": 0.0, "learning_rate": 0.00017704918032786885, "loss": 7.7031, "step": 27 }, { "epoch": 0.0, "learning_rate": 0.00018360655737704917, "loss": 7.4531, "step": 28 }, { "epoch": 0.0, "learning_rate": 0.0001901639344262295, "loss": 7.4805, "step": 29 }, { "epoch": 0.0, "learning_rate": 0.00019672131147540983, "loss": 7.1758, "step": 30 }, { "epoch": 0.0, "learning_rate": 0.00020327868852459017, "loss": 6.9258, "step": 31 }, { "epoch": 0.0, "learning_rate": 0.0002098360655737705, "loss": 6.7422, "step": 32 }, { "epoch": 0.0, "learning_rate": 0.00021639344262295082, "loss": 6.6406, "step": 33 }, { "epoch": 0.0, "learning_rate": 0.00022295081967213117, "loss": 6.4844, "step": 34 }, { "epoch": 0.0, "learning_rate": 0.00022950819672131148, "loss": 6.3906, "step": 35 }, { "epoch": 0.0, "learning_rate": 0.00023606557377049182, "loss": 6.3398, "step": 36 }, { "epoch": 0.0, "learning_rate": 0.00024262295081967214, "loss": 6.2266, "step": 37 }, { "epoch": 0.0, "learning_rate": 0.00024918032786885245, "loss": 6.0703, "step": 38 }, { "epoch": 0.0, "learning_rate": 0.00025573770491803277, "loss": 5.9492, "step": 39 }, { "epoch": 0.0, "learning_rate": 0.00026229508196721314, "loss": 5.9609, "step": 40 }, { "epoch": 0.0, "learning_rate": 0.00026885245901639345, "loss": 5.9023, "step": 41 }, { "epoch": 0.0, "learning_rate": 0.0002754098360655738, "loss": 5.8438, "step": 42 }, { "epoch": 0.0, "learning_rate": 0.0002819672131147541, "loss": 5.6992, "step": 43 }, { "epoch": 0.0, "learning_rate": 0.00028852459016393445, "loss": 5.6641, "step": 44 }, { "epoch": 0.0, "learning_rate": 0.00029508196721311476, "loss": 5.5664, "step": 45 }, { "epoch": 0.0, "learning_rate": 0.00030163934426229513, "loss": 5.5195, "step": 46 }, { "epoch": 0.0, "learning_rate": 0.0003081967213114754, "loss": 5.3477, "step": 47 }, { "epoch": 0.0, "learning_rate": 0.00031475409836065576, "loss": 5.2656, "step": 48 }, { "epoch": 0.0, "learning_rate": 0.0003213114754098361, "loss": 5.0742, "step": 49 }, { "epoch": 0.0, "learning_rate": 0.0003278688524590164, "loss": 4.9961, "step": 50 }, { "epoch": 0.01, "learning_rate": 0.0003344262295081967, "loss": 5.0352, "step": 51 }, { "epoch": 0.01, "learning_rate": 0.000340983606557377, "loss": 4.9141, "step": 52 }, { "epoch": 0.01, "learning_rate": 0.0003475409836065574, "loss": 4.8789, "step": 53 }, { "epoch": 0.01, "learning_rate": 0.0003540983606557377, "loss": 4.7773, "step": 54 }, { "epoch": 0.01, "learning_rate": 0.0003606557377049181, "loss": 4.6641, "step": 55 }, { "epoch": 0.01, "learning_rate": 0.00036721311475409834, "loss": 4.6992, "step": 56 }, { "epoch": 0.01, "learning_rate": 0.0003737704918032787, "loss": 4.5078, "step": 57 }, { "epoch": 0.01, "learning_rate": 0.000380327868852459, "loss": 4.5, "step": 58 }, { "epoch": 0.01, "learning_rate": 0.0003868852459016394, "loss": 4.2891, "step": 59 }, { "epoch": 0.01, "learning_rate": 0.00039344262295081965, "loss": 4.2344, "step": 60 }, { "epoch": 0.01, "learning_rate": 0.0004, "loss": 4.1758, "step": 61 }, { "epoch": 0.01, "learning_rate": 0.00040655737704918033, "loss": 4.0898, "step": 62 }, { "epoch": 0.01, "learning_rate": 0.0004131147540983607, "loss": 4.0781, "step": 63 }, { "epoch": 0.01, "learning_rate": 0.000419672131147541, "loss": 3.9785, "step": 64 }, { "epoch": 0.01, "learning_rate": 0.0004262295081967213, "loss": 3.9316, "step": 65 }, { "epoch": 0.01, "learning_rate": 0.00043278688524590165, "loss": 3.8672, "step": 66 }, { "epoch": 0.01, "learning_rate": 0.00043934426229508196, "loss": 3.7441, "step": 67 }, { "epoch": 0.01, "learning_rate": 0.00044590163934426233, "loss": 3.7422, "step": 68 }, { "epoch": 0.01, "learning_rate": 0.0004524590163934426, "loss": 3.5742, "step": 69 }, { "epoch": 0.01, "learning_rate": 0.00045901639344262296, "loss": 3.6602, "step": 70 }, { "epoch": 0.01, "learning_rate": 0.0004655737704918033, "loss": 3.5371, "step": 71 }, { "epoch": 0.01, "learning_rate": 0.00047213114754098365, "loss": 3.5723, "step": 72 }, { "epoch": 0.01, "learning_rate": 0.0004786885245901639, "loss": 3.4844, "step": 73 }, { "epoch": 0.01, "learning_rate": 0.0004852459016393443, "loss": 3.3457, "step": 74 }, { "epoch": 0.01, "learning_rate": 0.0004918032786885246, "loss": 3.4883, "step": 75 }, { "epoch": 0.01, "learning_rate": 0.0004983606557377049, "loss": 3.3281, "step": 76 }, { "epoch": 0.01, "learning_rate": 0.0005049180327868853, "loss": 3.1934, "step": 77 }, { "epoch": 0.01, "learning_rate": 0.0005114754098360655, "loss": 3.2637, "step": 78 }, { "epoch": 0.01, "learning_rate": 0.000518032786885246, "loss": 3.2109, "step": 79 }, { "epoch": 0.01, "learning_rate": 0.0005245901639344263, "loss": 3.2812, "step": 80 }, { "epoch": 0.01, "learning_rate": 0.0005311475409836065, "loss": 3.1602, "step": 81 }, { "epoch": 0.01, "learning_rate": 0.0005377049180327869, "loss": 3.2109, "step": 82 }, { "epoch": 0.01, "learning_rate": 0.0005442622950819672, "loss": 3.0996, "step": 83 }, { "epoch": 0.01, "learning_rate": 0.0005508196721311476, "loss": 3.0332, "step": 84 }, { "epoch": 0.01, "learning_rate": 0.0005573770491803278, "loss": 3.1465, "step": 85 }, { "epoch": 0.01, "learning_rate": 0.0005639344262295082, "loss": 3.0703, "step": 86 }, { "epoch": 0.01, "learning_rate": 0.0005704918032786886, "loss": 3.0664, "step": 87 }, { "epoch": 0.01, "learning_rate": 0.0005770491803278689, "loss": 2.9727, "step": 88 }, { "epoch": 0.01, "learning_rate": 0.0005836065573770491, "loss": 3.0195, "step": 89 }, { "epoch": 0.01, "learning_rate": 0.0005901639344262295, "loss": 2.9453, "step": 90 }, { "epoch": 0.01, "learning_rate": 0.0005967213114754098, "loss": 2.9238, "step": 91 }, { "epoch": 0.01, "learning_rate": 0.0006032786885245903, "loss": 2.8984, "step": 92 }, { "epoch": 0.01, "learning_rate": 0.0006098360655737705, "loss": 2.8516, "step": 93 }, { "epoch": 0.01, "learning_rate": 0.0006163934426229508, "loss": 2.9277, "step": 94 }, { "epoch": 0.01, "learning_rate": 0.0006229508196721312, "loss": 2.9238, "step": 95 }, { "epoch": 0.01, "learning_rate": 0.0006295081967213115, "loss": 2.8848, "step": 96 }, { "epoch": 0.01, "learning_rate": 0.0006360655737704918, "loss": 2.9375, "step": 97 }, { "epoch": 0.01, "learning_rate": 0.0006426229508196722, "loss": 2.9375, "step": 98 }, { "epoch": 0.01, "learning_rate": 0.0006491803278688525, "loss": 2.8047, "step": 99 }, { "epoch": 0.01, "learning_rate": 0.0006557377049180328, "loss": 2.9434, "step": 100 }, { "epoch": 0.01, "learning_rate": 0.0006622950819672131, "loss": 2.7637, "step": 101 }, { "epoch": 0.01, "learning_rate": 0.0006688524590163934, "loss": 2.8398, "step": 102 }, { "epoch": 0.01, "learning_rate": 0.0006754098360655738, "loss": 2.832, "step": 103 }, { "epoch": 0.01, "learning_rate": 0.000681967213114754, "loss": 2.8535, "step": 104 }, { "epoch": 0.01, "learning_rate": 0.0006885245901639345, "loss": 2.8418, "step": 105 }, { "epoch": 0.01, "learning_rate": 0.0006950819672131148, "loss": 2.7148, "step": 106 }, { "epoch": 0.01, "learning_rate": 0.0007016393442622951, "loss": 2.6895, "step": 107 }, { "epoch": 0.01, "learning_rate": 0.0007081967213114754, "loss": 2.7637, "step": 108 }, { "epoch": 0.01, "learning_rate": 0.0007147540983606557, "loss": 2.8691, "step": 109 }, { "epoch": 0.01, "learning_rate": 0.0007213114754098362, "loss": 2.7676, "step": 110 }, { "epoch": 0.01, "learning_rate": 0.0007278688524590165, "loss": 2.6816, "step": 111 }, { "epoch": 0.01, "learning_rate": 0.0007344262295081967, "loss": 2.6543, "step": 112 }, { "epoch": 0.01, "learning_rate": 0.0007409836065573771, "loss": 2.6328, "step": 113 }, { "epoch": 0.01, "learning_rate": 0.0007475409836065574, "loss": 2.6855, "step": 114 }, { "epoch": 0.01, "learning_rate": 0.0007540983606557376, "loss": 2.6777, "step": 115 }, { "epoch": 0.01, "learning_rate": 0.000760655737704918, "loss": 2.7168, "step": 116 }, { "epoch": 0.01, "learning_rate": 0.0007672131147540984, "loss": 2.6387, "step": 117 }, { "epoch": 0.01, "learning_rate": 0.0007737704918032788, "loss": 2.6973, "step": 118 }, { "epoch": 0.01, "learning_rate": 0.000780327868852459, "loss": 2.6836, "step": 119 }, { "epoch": 0.01, "learning_rate": 0.0007868852459016393, "loss": 2.623, "step": 120 }, { "epoch": 0.01, "learning_rate": 0.0007934426229508197, "loss": 2.7188, "step": 121 }, { "epoch": 0.01, "learning_rate": 0.0008, "loss": 2.5938, "step": 122 }, { "epoch": 0.01, "learning_rate": 0.0008065573770491804, "loss": 2.6348, "step": 123 }, { "epoch": 0.01, "learning_rate": 0.0008131147540983607, "loss": 2.5957, "step": 124 }, { "epoch": 0.01, "learning_rate": 0.000819672131147541, "loss": 2.5742, "step": 125 }, { "epoch": 0.01, "learning_rate": 0.0008262295081967214, "loss": 2.6113, "step": 126 }, { "epoch": 0.01, "learning_rate": 0.0008327868852459016, "loss": 2.625, "step": 127 }, { "epoch": 0.01, "learning_rate": 0.000839344262295082, "loss": 2.5703, "step": 128 }, { "epoch": 0.01, "learning_rate": 0.0008459016393442624, "loss": 2.6816, "step": 129 }, { "epoch": 0.01, "learning_rate": 0.0008524590163934426, "loss": 2.5703, "step": 130 }, { "epoch": 0.01, "learning_rate": 0.000859016393442623, "loss": 2.5703, "step": 131 }, { "epoch": 0.01, "learning_rate": 0.0008655737704918033, "loss": 2.584, "step": 132 }, { "epoch": 0.01, "learning_rate": 0.0008721311475409836, "loss": 2.5488, "step": 133 }, { "epoch": 0.01, "learning_rate": 0.0008786885245901639, "loss": 2.4922, "step": 134 }, { "epoch": 0.01, "learning_rate": 0.0008852459016393442, "loss": 2.4941, "step": 135 }, { "epoch": 0.01, "learning_rate": 0.0008918032786885247, "loss": 2.5117, "step": 136 }, { "epoch": 0.01, "learning_rate": 0.000898360655737705, "loss": 2.4941, "step": 137 }, { "epoch": 0.01, "learning_rate": 0.0009049180327868852, "loss": 2.5762, "step": 138 }, { "epoch": 0.01, "learning_rate": 0.0009114754098360656, "loss": 2.6172, "step": 139 }, { "epoch": 0.01, "learning_rate": 0.0009180327868852459, "loss": 2.4844, "step": 140 }, { "epoch": 0.01, "learning_rate": 0.0009245901639344263, "loss": 2.4941, "step": 141 }, { "epoch": 0.01, "learning_rate": 0.0009311475409836066, "loss": 2.5059, "step": 142 }, { "epoch": 0.01, "learning_rate": 0.0009377049180327869, "loss": 2.6309, "step": 143 }, { "epoch": 0.01, "learning_rate": 0.0009442622950819673, "loss": 2.502, "step": 144 }, { "epoch": 0.01, "learning_rate": 0.0009508196721311475, "loss": 2.5605, "step": 145 }, { "epoch": 0.01, "learning_rate": 0.0009573770491803278, "loss": 2.4805, "step": 146 }, { "epoch": 0.01, "learning_rate": 0.0009639344262295082, "loss": 2.5508, "step": 147 }, { "epoch": 0.01, "learning_rate": 0.0009704918032786885, "loss": 2.584, "step": 148 }, { "epoch": 0.01, "learning_rate": 0.0009770491803278688, "loss": 2.5703, "step": 149 }, { "epoch": 0.01, "learning_rate": 0.0009836065573770492, "loss": 2.4902, "step": 150 }, { "epoch": 0.01, "learning_rate": 0.0009901639344262296, "loss": 2.5176, "step": 151 }, { "epoch": 0.01, "learning_rate": 0.0009967213114754098, "loss": 2.4902, "step": 152 }, { "epoch": 0.02, "learning_rate": 0.0010032786885245902, "loss": 2.4648, "step": 153 }, { "epoch": 0.02, "learning_rate": 0.0010098360655737707, "loss": 2.4453, "step": 154 }, { "epoch": 0.02, "learning_rate": 0.0010163934426229509, "loss": 2.4219, "step": 155 }, { "epoch": 0.02, "learning_rate": 0.001022950819672131, "loss": 2.459, "step": 156 }, { "epoch": 0.02, "learning_rate": 0.0010295081967213115, "loss": 2.4023, "step": 157 }, { "epoch": 0.02, "learning_rate": 0.001036065573770492, "loss": 2.4492, "step": 158 }, { "epoch": 0.02, "learning_rate": 0.0010426229508196721, "loss": 2.5137, "step": 159 }, { "epoch": 0.02, "learning_rate": 0.0010491803278688525, "loss": 2.459, "step": 160 }, { "epoch": 0.02, "learning_rate": 0.0010557377049180328, "loss": 2.4258, "step": 161 }, { "epoch": 0.02, "learning_rate": 0.001062295081967213, "loss": 2.3789, "step": 162 }, { "epoch": 0.02, "learning_rate": 0.0010688524590163936, "loss": 2.5352, "step": 163 }, { "epoch": 0.02, "learning_rate": 0.0010754098360655738, "loss": 2.4141, "step": 164 }, { "epoch": 0.02, "learning_rate": 0.0010819672131147542, "loss": 2.5059, "step": 165 }, { "epoch": 0.02, "learning_rate": 0.0010885245901639344, "loss": 2.4102, "step": 166 }, { "epoch": 0.02, "learning_rate": 0.0010950819672131146, "loss": 2.4648, "step": 167 }, { "epoch": 0.02, "learning_rate": 0.0011016393442622953, "loss": 2.4102, "step": 168 }, { "epoch": 0.02, "learning_rate": 0.0011081967213114755, "loss": 2.4141, "step": 169 }, { "epoch": 0.02, "learning_rate": 0.0011147540983606557, "loss": 2.4336, "step": 170 }, { "epoch": 0.02, "learning_rate": 0.0011213114754098361, "loss": 2.4922, "step": 171 }, { "epoch": 0.02, "learning_rate": 0.0011278688524590163, "loss": 2.375, "step": 172 }, { "epoch": 0.02, "learning_rate": 0.0011344262295081967, "loss": 2.3945, "step": 173 }, { "epoch": 0.02, "learning_rate": 0.0011409836065573772, "loss": 2.3887, "step": 174 }, { "epoch": 0.02, "learning_rate": 0.0011475409836065574, "loss": 2.4922, "step": 175 }, { "epoch": 0.02, "learning_rate": 0.0011540983606557378, "loss": 2.4727, "step": 176 }, { "epoch": 0.02, "learning_rate": 0.001160655737704918, "loss": 2.4297, "step": 177 }, { "epoch": 0.02, "learning_rate": 0.0011672131147540982, "loss": 2.3965, "step": 178 }, { "epoch": 0.02, "learning_rate": 0.0011737704918032789, "loss": 2.4629, "step": 179 }, { "epoch": 0.02, "learning_rate": 0.001180327868852459, "loss": 2.373, "step": 180 }, { "epoch": 0.02, "learning_rate": 0.0011868852459016393, "loss": 2.3965, "step": 181 }, { "epoch": 0.02, "learning_rate": 0.0011934426229508197, "loss": 2.3926, "step": 182 }, { "epoch": 0.02, "learning_rate": 0.0012, "loss": 2.3926, "step": 183 }, { "epoch": 0.02, "learning_rate": 0.0012065573770491805, "loss": 2.4219, "step": 184 }, { "epoch": 0.02, "learning_rate": 0.0012131147540983607, "loss": 2.3945, "step": 185 }, { "epoch": 0.02, "learning_rate": 0.001219672131147541, "loss": 2.3477, "step": 186 }, { "epoch": 0.02, "learning_rate": 0.0012262295081967214, "loss": 2.418, "step": 187 }, { "epoch": 0.02, "learning_rate": 0.0012327868852459016, "loss": 2.4297, "step": 188 }, { "epoch": 0.02, "learning_rate": 0.001239344262295082, "loss": 2.3203, "step": 189 }, { "epoch": 0.02, "learning_rate": 0.0012459016393442624, "loss": 2.4023, "step": 190 }, { "epoch": 0.02, "learning_rate": 0.0012524590163934426, "loss": 2.3789, "step": 191 }, { "epoch": 0.02, "learning_rate": 0.001259016393442623, "loss": 2.3887, "step": 192 }, { "epoch": 0.02, "learning_rate": 0.0012655737704918033, "loss": 2.3945, "step": 193 }, { "epoch": 0.02, "learning_rate": 0.0012721311475409837, "loss": 2.4121, "step": 194 }, { "epoch": 0.02, "learning_rate": 0.001278688524590164, "loss": 2.373, "step": 195 }, { "epoch": 0.02, "learning_rate": 0.0012852459016393443, "loss": 2.3535, "step": 196 }, { "epoch": 0.02, "learning_rate": 0.0012918032786885245, "loss": 2.4023, "step": 197 }, { "epoch": 0.02, "learning_rate": 0.001298360655737705, "loss": 2.3301, "step": 198 }, { "epoch": 0.02, "learning_rate": 0.0013049180327868851, "loss": 2.541, "step": 199 }, { "epoch": 0.02, "learning_rate": 0.0013114754098360656, "loss": 2.4102, "step": 200 }, { "epoch": 0.02, "learning_rate": 0.001318032786885246, "loss": 2.2734, "step": 201 }, { "epoch": 0.02, "learning_rate": 0.0013245901639344262, "loss": 2.3535, "step": 202 }, { "epoch": 0.02, "learning_rate": 0.0013311475409836066, "loss": 2.3496, "step": 203 }, { "epoch": 0.02, "learning_rate": 0.0013377049180327868, "loss": 2.3184, "step": 204 }, { "epoch": 0.02, "learning_rate": 0.0013442622950819673, "loss": 2.332, "step": 205 }, { "epoch": 0.02, "learning_rate": 0.0013508196721311477, "loss": 2.3203, "step": 206 }, { "epoch": 0.02, "learning_rate": 0.0013573770491803279, "loss": 2.3945, "step": 207 }, { "epoch": 0.02, "learning_rate": 0.001363934426229508, "loss": 2.4102, "step": 208 }, { "epoch": 0.02, "learning_rate": 0.0013704918032786885, "loss": 2.3594, "step": 209 }, { "epoch": 0.02, "learning_rate": 0.001377049180327869, "loss": 2.3613, "step": 210 }, { "epoch": 0.02, "learning_rate": 0.0013836065573770491, "loss": 2.4336, "step": 211 }, { "epoch": 0.02, "learning_rate": 0.0013901639344262296, "loss": 2.3965, "step": 212 }, { "epoch": 0.02, "learning_rate": 0.0013967213114754098, "loss": 2.3184, "step": 213 }, { "epoch": 0.02, "learning_rate": 0.0014032786885245902, "loss": 2.375, "step": 214 }, { "epoch": 0.02, "learning_rate": 0.0014098360655737706, "loss": 2.2285, "step": 215 }, { "epoch": 0.02, "learning_rate": 0.0014163934426229508, "loss": 2.3633, "step": 216 }, { "epoch": 0.02, "learning_rate": 0.0014229508196721313, "loss": 2.2871, "step": 217 }, { "epoch": 0.02, "learning_rate": 0.0014295081967213115, "loss": 2.2871, "step": 218 }, { "epoch": 0.02, "learning_rate": 0.0014360655737704917, "loss": 2.3125, "step": 219 }, { "epoch": 0.02, "learning_rate": 0.0014426229508196723, "loss": 2.2852, "step": 220 }, { "epoch": 0.02, "learning_rate": 0.0014491803278688525, "loss": 2.3711, "step": 221 }, { "epoch": 0.02, "learning_rate": 0.001455737704918033, "loss": 2.3066, "step": 222 }, { "epoch": 0.02, "learning_rate": 0.0014622950819672131, "loss": 2.3496, "step": 223 }, { "epoch": 0.02, "learning_rate": 0.0014688524590163933, "loss": 2.3125, "step": 224 }, { "epoch": 0.02, "learning_rate": 0.001475409836065574, "loss": 2.3828, "step": 225 }, { "epoch": 0.02, "learning_rate": 0.0014819672131147542, "loss": 2.2715, "step": 226 }, { "epoch": 0.02, "learning_rate": 0.0014885245901639344, "loss": 2.4199, "step": 227 }, { "epoch": 0.02, "learning_rate": 0.0014950819672131148, "loss": 2.3887, "step": 228 }, { "epoch": 0.02, "learning_rate": 0.001501639344262295, "loss": 2.332, "step": 229 }, { "epoch": 0.02, "learning_rate": 0.0015081967213114752, "loss": 2.2559, "step": 230 }, { "epoch": 0.02, "learning_rate": 0.0015147540983606559, "loss": 2.3164, "step": 231 }, { "epoch": 0.02, "learning_rate": 0.001521311475409836, "loss": 2.2891, "step": 232 }, { "epoch": 0.02, "learning_rate": 0.0015278688524590165, "loss": 2.3398, "step": 233 }, { "epoch": 0.02, "learning_rate": 0.0015344262295081967, "loss": 2.3105, "step": 234 }, { "epoch": 0.02, "learning_rate": 0.001540983606557377, "loss": 2.3066, "step": 235 }, { "epoch": 0.02, "learning_rate": 0.0015475409836065576, "loss": 2.3262, "step": 236 }, { "epoch": 0.02, "learning_rate": 0.0015540983606557378, "loss": 2.2266, "step": 237 }, { "epoch": 0.02, "learning_rate": 0.001560655737704918, "loss": 2.375, "step": 238 }, { "epoch": 0.02, "learning_rate": 0.0015672131147540984, "loss": 2.2969, "step": 239 }, { "epoch": 0.02, "learning_rate": 0.0015737704918032786, "loss": 2.3184, "step": 240 }, { "epoch": 0.02, "learning_rate": 0.001580327868852459, "loss": 2.2793, "step": 241 }, { "epoch": 0.02, "learning_rate": 0.0015868852459016394, "loss": 2.3691, "step": 242 }, { "epoch": 0.02, "learning_rate": 0.0015934426229508197, "loss": 2.2441, "step": 243 }, { "epoch": 0.02, "learning_rate": 0.0016, "loss": 2.2305, "step": 244 }, { "epoch": 0.02, "learning_rate": 0.0016065573770491803, "loss": 2.3496, "step": 245 }, { "epoch": 0.02, "learning_rate": 0.0016131147540983607, "loss": 2.2148, "step": 246 }, { "epoch": 0.02, "learning_rate": 0.0016196721311475411, "loss": 2.3848, "step": 247 }, { "epoch": 0.02, "learning_rate": 0.0016262295081967213, "loss": 2.2988, "step": 248 }, { "epoch": 0.02, "learning_rate": 0.0016327868852459015, "loss": 2.2852, "step": 249 }, { "epoch": 0.02, "learning_rate": 0.001639344262295082, "loss": 2.2168, "step": 250 }, { "epoch": 0.02, "learning_rate": 0.0016459016393442624, "loss": 2.293, "step": 251 }, { "epoch": 0.02, "learning_rate": 0.0016524590163934428, "loss": 2.3613, "step": 252 }, { "epoch": 0.02, "learning_rate": 0.001659016393442623, "loss": 2.2793, "step": 253 }, { "epoch": 0.03, "learning_rate": 0.0016655737704918032, "loss": 2.208, "step": 254 }, { "epoch": 0.03, "learning_rate": 0.0016721311475409836, "loss": 2.2266, "step": 255 }, { "epoch": 0.03, "learning_rate": 0.001678688524590164, "loss": 2.2578, "step": 256 }, { "epoch": 0.03, "learning_rate": 0.0016852459016393443, "loss": 2.2754, "step": 257 }, { "epoch": 0.03, "learning_rate": 0.0016918032786885247, "loss": 2.2734, "step": 258 }, { "epoch": 0.03, "learning_rate": 0.001698360655737705, "loss": 2.2715, "step": 259 }, { "epoch": 0.03, "learning_rate": 0.0017049180327868851, "loss": 2.2773, "step": 260 }, { "epoch": 0.03, "learning_rate": 0.0017114754098360655, "loss": 2.2363, "step": 261 }, { "epoch": 0.03, "learning_rate": 0.001718032786885246, "loss": 2.2549, "step": 262 }, { "epoch": 0.03, "learning_rate": 0.0017245901639344264, "loss": 2.2949, "step": 263 }, { "epoch": 0.03, "learning_rate": 0.0017311475409836066, "loss": 2.2178, "step": 264 }, { "epoch": 0.03, "learning_rate": 0.0017377049180327868, "loss": 2.3281, "step": 265 }, { "epoch": 0.03, "learning_rate": 0.0017442622950819672, "loss": 2.2578, "step": 266 }, { "epoch": 0.03, "learning_rate": 0.0017508196721311476, "loss": 2.2852, "step": 267 }, { "epoch": 0.03, "learning_rate": 0.0017573770491803278, "loss": 2.3242, "step": 268 }, { "epoch": 0.03, "learning_rate": 0.0017639344262295083, "loss": 2.3184, "step": 269 }, { "epoch": 0.03, "learning_rate": 0.0017704918032786885, "loss": 2.1709, "step": 270 }, { "epoch": 0.03, "learning_rate": 0.001777049180327869, "loss": 2.252, "step": 271 }, { "epoch": 0.03, "learning_rate": 0.0017836065573770493, "loss": 2.2754, "step": 272 }, { "epoch": 0.03, "learning_rate": 0.0017901639344262295, "loss": 2.2324, "step": 273 }, { "epoch": 0.03, "learning_rate": 0.00179672131147541, "loss": 2.2168, "step": 274 }, { "epoch": 0.03, "learning_rate": 0.0018032786885245902, "loss": 2.3379, "step": 275 }, { "epoch": 0.03, "learning_rate": 0.0018098360655737704, "loss": 2.2266, "step": 276 }, { "epoch": 0.03, "learning_rate": 0.001816393442622951, "loss": 2.1904, "step": 277 }, { "epoch": 0.03, "learning_rate": 0.0018229508196721312, "loss": 2.2891, "step": 278 }, { "epoch": 0.03, "learning_rate": 0.0018295081967213114, "loss": 2.2412, "step": 279 }, { "epoch": 0.03, "learning_rate": 0.0018360655737704918, "loss": 2.1797, "step": 280 }, { "epoch": 0.03, "learning_rate": 0.001842622950819672, "loss": 2.2305, "step": 281 }, { "epoch": 0.03, "learning_rate": 0.0018491803278688527, "loss": 2.3828, "step": 282 }, { "epoch": 0.03, "learning_rate": 0.001855737704918033, "loss": 2.2637, "step": 283 }, { "epoch": 0.03, "learning_rate": 0.001862295081967213, "loss": 2.252, "step": 284 }, { "epoch": 0.03, "learning_rate": 0.0018688524590163935, "loss": 2.1973, "step": 285 }, { "epoch": 0.03, "learning_rate": 0.0018754098360655737, "loss": 2.2578, "step": 286 }, { "epoch": 0.03, "learning_rate": 0.0018819672131147542, "loss": 2.3174, "step": 287 }, { "epoch": 0.03, "learning_rate": 0.0018885245901639346, "loss": 2.2754, "step": 288 }, { "epoch": 0.03, "learning_rate": 0.0018950819672131148, "loss": 2.2822, "step": 289 }, { "epoch": 0.03, "learning_rate": 0.001901639344262295, "loss": 2.207, "step": 290 }, { "epoch": 0.03, "learning_rate": 0.0019081967213114754, "loss": 2.2891, "step": 291 }, { "epoch": 0.03, "learning_rate": 0.0019147540983606556, "loss": 2.1914, "step": 292 }, { "epoch": 0.03, "learning_rate": 0.0019213114754098363, "loss": 2.2422, "step": 293 }, { "epoch": 0.03, "learning_rate": 0.0019278688524590165, "loss": 2.2354, "step": 294 }, { "epoch": 0.03, "learning_rate": 0.0019344262295081967, "loss": 2.209, "step": 295 }, { "epoch": 0.03, "learning_rate": 0.001940983606557377, "loss": 2.3223, "step": 296 }, { "epoch": 0.03, "learning_rate": 0.0019475409836065573, "loss": 2.2334, "step": 297 }, { "epoch": 0.03, "learning_rate": 0.0019540983606557375, "loss": 2.2871, "step": 298 }, { "epoch": 0.03, "learning_rate": 0.001960655737704918, "loss": 2.1504, "step": 299 }, { "epoch": 0.03, "learning_rate": 0.0019672131147540984, "loss": 2.2754, "step": 300 }, { "epoch": 0.03, "learning_rate": 0.0019737704918032788, "loss": 2.2188, "step": 301 }, { "epoch": 0.03, "learning_rate": 0.001980327868852459, "loss": 2.2012, "step": 302 }, { "epoch": 0.03, "learning_rate": 0.001986885245901639, "loss": 2.2656, "step": 303 }, { "epoch": 0.03, "learning_rate": 0.0019934426229508196, "loss": 2.2383, "step": 304 }, { "epoch": 0.03, "learning_rate": 0.002, "loss": 2.2559, "step": 305 }, { "epoch": 0.03, "learning_rate": 0.001999999949023757, "loss": 2.2227, "step": 306 }, { "epoch": 0.03, "learning_rate": 0.001999999796095033, "loss": 2.2695, "step": 307 }, { "epoch": 0.03, "learning_rate": 0.001999999541213844, "loss": 2.2266, "step": 308 }, { "epoch": 0.03, "learning_rate": 0.001999999184380215, "loss": 2.126, "step": 309 }, { "epoch": 0.03, "learning_rate": 0.0019999987255941835, "loss": 2.3047, "step": 310 }, { "epoch": 0.03, "learning_rate": 0.0019999981648557955, "loss": 2.2227, "step": 311 }, { "epoch": 0.03, "learning_rate": 0.001999997502165109, "loss": 2.2461, "step": 312 }, { "epoch": 0.03, "learning_rate": 0.0019999967375221904, "loss": 2.1699, "step": 313 }, { "epoch": 0.03, "learning_rate": 0.0019999958709271187, "loss": 2.1973, "step": 314 }, { "epoch": 0.03, "learning_rate": 0.0019999949023799817, "loss": 2.1182, "step": 315 }, { "epoch": 0.03, "learning_rate": 0.0019999938318808782, "loss": 2.2354, "step": 316 }, { "epoch": 0.03, "learning_rate": 0.0019999926594299177, "loss": 2.2441, "step": 317 }, { "epoch": 0.03, "learning_rate": 0.0019999913850272193, "loss": 2.2324, "step": 318 }, { "epoch": 0.03, "learning_rate": 0.001999990008672913, "loss": 2.2266, "step": 319 }, { "epoch": 0.03, "learning_rate": 0.0019999885303671394, "loss": 2.1855, "step": 320 }, { "epoch": 0.03, "learning_rate": 0.001999986950110049, "loss": 2.2246, "step": 321 }, { "epoch": 0.03, "learning_rate": 0.001999985267901803, "loss": 2.2061, "step": 322 }, { "epoch": 0.03, "learning_rate": 0.0019999834837425724, "loss": 2.2402, "step": 323 }, { "epoch": 0.03, "learning_rate": 0.0019999815976325403, "loss": 2.1279, "step": 324 }, { "epoch": 0.03, "learning_rate": 0.001999979609571898, "loss": 2.2363, "step": 325 }, { "epoch": 0.03, "learning_rate": 0.001999977519560848, "loss": 2.1641, "step": 326 }, { "epoch": 0.03, "learning_rate": 0.0019999753275996043, "loss": 2.2188, "step": 327 }, { "epoch": 0.03, "learning_rate": 0.0019999730336883897, "loss": 2.1865, "step": 328 }, { "epoch": 0.03, "learning_rate": 0.0019999706378274382, "loss": 2.168, "step": 329 }, { "epoch": 0.03, "learning_rate": 0.001999968140016994, "loss": 2.2793, "step": 330 }, { "epoch": 0.03, "learning_rate": 0.0019999655402573122, "loss": 2.2266, "step": 331 }, { "epoch": 0.03, "learning_rate": 0.0019999628385486577, "loss": 2.3105, "step": 332 }, { "epoch": 0.03, "learning_rate": 0.001999960034891305, "loss": 2.2422, "step": 333 }, { "epoch": 0.03, "learning_rate": 0.0019999571292855416, "loss": 2.2344, "step": 334 }, { "epoch": 0.03, "learning_rate": 0.0019999541217316625, "loss": 2.1416, "step": 335 }, { "epoch": 0.03, "learning_rate": 0.0019999510122299747, "loss": 2.1523, "step": 336 }, { "epoch": 0.03, "learning_rate": 0.001999947800780795, "loss": 2.1846, "step": 337 }, { "epoch": 0.03, "learning_rate": 0.0019999444873844514, "loss": 2.2061, "step": 338 }, { "epoch": 0.03, "learning_rate": 0.001999941072041281, "loss": 2.1465, "step": 339 }, { "epoch": 0.03, "learning_rate": 0.0019999375547516325, "loss": 2.2559, "step": 340 }, { "epoch": 0.03, "learning_rate": 0.001999933935515864, "loss": 2.1963, "step": 341 }, { "epoch": 0.03, "learning_rate": 0.001999930214334345, "loss": 2.252, "step": 342 }, { "epoch": 0.03, "learning_rate": 0.001999926391207455, "loss": 2.1543, "step": 343 }, { "epoch": 0.03, "learning_rate": 0.001999922466135583, "loss": 2.2051, "step": 344 }, { "epoch": 0.03, "learning_rate": 0.00199991843911913, "loss": 2.248, "step": 345 }, { "epoch": 0.03, "learning_rate": 0.0019999143101585063, "loss": 2.2207, "step": 346 }, { "epoch": 0.03, "learning_rate": 0.0019999100792541325, "loss": 2.1523, "step": 347 }, { "epoch": 0.03, "learning_rate": 0.0019999057464064403, "loss": 2.2324, "step": 348 }, { "epoch": 0.03, "learning_rate": 0.0019999013116158717, "loss": 2.2383, "step": 349 }, { "epoch": 0.03, "learning_rate": 0.001999896774882878, "loss": 2.291, "step": 350 }, { "epoch": 0.03, "learning_rate": 0.001999892136207922, "loss": 2.2637, "step": 351 }, { "epoch": 0.03, "learning_rate": 0.0019998873955914776, "loss": 2.2754, "step": 352 }, { "epoch": 0.03, "learning_rate": 0.001999882553034027, "loss": 2.2051, "step": 353 }, { "epoch": 0.03, "learning_rate": 0.0019998776085360646, "loss": 2.1797, "step": 354 }, { "epoch": 0.03, "learning_rate": 0.001999872562098094, "loss": 2.1719, "step": 355 }, { "epoch": 0.04, "learning_rate": 0.0019998674137206296, "loss": 2.1592, "step": 356 }, { "epoch": 0.04, "learning_rate": 0.001999862163404197, "loss": 2.209, "step": 357 }, { "epoch": 0.04, "learning_rate": 0.001999856811149331, "loss": 2.0693, "step": 358 }, { "epoch": 0.04, "learning_rate": 0.0019998513569565767, "loss": 2.2051, "step": 359 }, { "epoch": 0.04, "learning_rate": 0.0019998458008264916, "loss": 2.1875, "step": 360 }, { "epoch": 0.04, "learning_rate": 0.001999840142759641, "loss": 2.2422, "step": 361 }, { "epoch": 0.04, "learning_rate": 0.0019998343827566023, "loss": 2.293, "step": 362 }, { "epoch": 0.04, "learning_rate": 0.0019998285208179625, "loss": 2.2285, "step": 363 }, { "epoch": 0.04, "learning_rate": 0.0019998225569443193, "loss": 2.2119, "step": 364 }, { "epoch": 0.04, "learning_rate": 0.0019998164911362807, "loss": 2.085, "step": 365 }, { "epoch": 0.04, "learning_rate": 0.001999810323394465, "loss": 2.1553, "step": 366 }, { "epoch": 0.04, "learning_rate": 0.0019998040537195015, "loss": 2.166, "step": 367 }, { "epoch": 0.04, "learning_rate": 0.001999797682112029, "loss": 2.1553, "step": 368 }, { "epoch": 0.04, "learning_rate": 0.001999791208572697, "loss": 2.1699, "step": 369 }, { "epoch": 0.04, "learning_rate": 0.001999784633102166, "loss": 2.248, "step": 370 }, { "epoch": 0.04, "learning_rate": 0.001999777955701106, "loss": 2.1787, "step": 371 }, { "epoch": 0.04, "learning_rate": 0.0019997711763701975, "loss": 2.1113, "step": 372 }, { "epoch": 0.04, "learning_rate": 0.0019997642951101323, "loss": 2.1787, "step": 373 }, { "epoch": 0.04, "learning_rate": 0.0019997573119216114, "loss": 2.2324, "step": 374 }, { "epoch": 0.04, "learning_rate": 0.0019997502268053477, "loss": 2.1484, "step": 375 }, { "epoch": 0.04, "learning_rate": 0.001999743039762062, "loss": 2.1602, "step": 376 }, { "epoch": 0.04, "learning_rate": 0.0019997357507924884, "loss": 2.0225, "step": 377 }, { "epoch": 0.04, "learning_rate": 0.00199972835989737, "loss": 2.2031, "step": 378 }, { "epoch": 0.04, "learning_rate": 0.0019997208670774593, "loss": 2.1465, "step": 379 }, { "epoch": 0.04, "learning_rate": 0.0019997132723335205, "loss": 2.1543, "step": 380 }, { "epoch": 0.04, "learning_rate": 0.0019997055756663285, "loss": 2.1025, "step": 381 }, { "epoch": 0.04, "learning_rate": 0.0019996977770766677, "loss": 2.1543, "step": 382 }, { "epoch": 0.04, "learning_rate": 0.001999689876565333, "loss": 2.1406, "step": 383 }, { "epoch": 0.04, "learning_rate": 0.0019996818741331306, "loss": 2.1504, "step": 384 }, { "epoch": 0.04, "learning_rate": 0.0019996737697808753, "loss": 2.1553, "step": 385 }, { "epoch": 0.04, "learning_rate": 0.0019996655635093935, "loss": 2.1025, "step": 386 }, { "epoch": 0.04, "learning_rate": 0.001999657255319523, "loss": 2.0898, "step": 387 }, { "epoch": 0.04, "learning_rate": 0.0019996488452121096, "loss": 2.1152, "step": 388 }, { "epoch": 0.04, "learning_rate": 0.0019996403331880115, "loss": 2.1465, "step": 389 }, { "epoch": 0.04, "learning_rate": 0.0019996317192480955, "loss": 2.1133, "step": 390 }, { "epoch": 0.04, "learning_rate": 0.0019996230033932413, "loss": 2.1465, "step": 391 }, { "epoch": 0.04, "learning_rate": 0.0019996141856243365, "loss": 2.2656, "step": 392 }, { "epoch": 0.04, "learning_rate": 0.00199960526594228, "loss": 2.1211, "step": 393 }, { "epoch": 0.04, "learning_rate": 0.0019995962443479817, "loss": 2.1211, "step": 394 }, { "epoch": 0.04, "learning_rate": 0.0019995871208423614, "loss": 2.1211, "step": 395 }, { "epoch": 0.04, "learning_rate": 0.0019995778954263487, "loss": 2.1689, "step": 396 }, { "epoch": 0.04, "learning_rate": 0.0019995685681008847, "loss": 2.1025, "step": 397 }, { "epoch": 0.04, "learning_rate": 0.0019995591388669203, "loss": 2.0723, "step": 398 }, { "epoch": 0.04, "learning_rate": 0.001999549607725416, "loss": 2.1523, "step": 399 }, { "epoch": 0.04, "learning_rate": 0.0019995399746773447, "loss": 2.1299, "step": 400 }, { "epoch": 0.04, "learning_rate": 0.001999530239723688, "loss": 2.1709, "step": 401 }, { "epoch": 0.04, "learning_rate": 0.0019995204028654383, "loss": 2.1709, "step": 402 }, { "epoch": 0.04, "learning_rate": 0.001999510464103599, "loss": 2.2021, "step": 403 }, { "epoch": 0.04, "learning_rate": 0.001999500423439182, "loss": 2.1387, "step": 404 }, { "epoch": 0.04, "learning_rate": 0.0019994902808732127, "loss": 2.166, "step": 405 }, { "epoch": 0.04, "learning_rate": 0.0019994800364067245, "loss": 2.1504, "step": 406 }, { "epoch": 0.04, "learning_rate": 0.0019994696900407613, "loss": 2.1924, "step": 407 }, { "epoch": 0.04, "learning_rate": 0.0019994592417763787, "loss": 2.126, "step": 408 }, { "epoch": 0.04, "learning_rate": 0.001999448691614642, "loss": 2.1025, "step": 409 }, { "epoch": 0.04, "learning_rate": 0.0019994380395566254, "loss": 2.2031, "step": 410 }, { "epoch": 0.04, "learning_rate": 0.0019994272856034166, "loss": 2.1494, "step": 411 }, { "epoch": 0.04, "learning_rate": 0.0019994164297561113, "loss": 2.1934, "step": 412 }, { "epoch": 0.04, "learning_rate": 0.0019994054720158163, "loss": 2.1611, "step": 413 }, { "epoch": 0.04, "learning_rate": 0.0019993944123836488, "loss": 2.1562, "step": 414 }, { "epoch": 0.04, "learning_rate": 0.0019993832508607362, "loss": 2.1689, "step": 415 }, { "epoch": 0.04, "learning_rate": 0.0019993719874482167, "loss": 2.2598, "step": 416 }, { "epoch": 0.04, "learning_rate": 0.0019993606221472386, "loss": 2.2451, "step": 417 }, { "epoch": 0.04, "learning_rate": 0.0019993491549589603, "loss": 2.1006, "step": 418 }, { "epoch": 0.04, "learning_rate": 0.0019993375858845517, "loss": 2.0781, "step": 419 }, { "epoch": 0.04, "learning_rate": 0.001999325914925191, "loss": 2.1279, "step": 420 }, { "epoch": 0.04, "learning_rate": 0.0019993141420820695, "loss": 2.1875, "step": 421 }, { "epoch": 0.04, "learning_rate": 0.0019993022673563863, "loss": 2.1113, "step": 422 }, { "epoch": 0.04, "learning_rate": 0.0019992902907493525, "loss": 2.1973, "step": 423 }, { "epoch": 0.04, "learning_rate": 0.0019992782122621893, "loss": 2.1914, "step": 424 }, { "epoch": 0.04, "learning_rate": 0.0019992660318961285, "loss": 1.9873, "step": 425 }, { "epoch": 0.04, "learning_rate": 0.001999253749652411, "loss": 2.1133, "step": 426 }, { "epoch": 0.04, "learning_rate": 0.0019992413655322895, "loss": 2.1348, "step": 427 }, { "epoch": 0.04, "learning_rate": 0.001999228879537027, "loss": 2.1719, "step": 428 }, { "epoch": 0.04, "learning_rate": 0.0019992162916678954, "loss": 2.1621, "step": 429 }, { "epoch": 0.04, "learning_rate": 0.001999203601926179, "loss": 2.1855, "step": 430 }, { "epoch": 0.04, "learning_rate": 0.0019991908103131712, "loss": 2.1455, "step": 431 }, { "epoch": 0.04, "learning_rate": 0.001999177916830176, "loss": 2.084, "step": 432 }, { "epoch": 0.04, "learning_rate": 0.001999164921478508, "loss": 2.1064, "step": 433 }, { "epoch": 0.04, "learning_rate": 0.001999151824259493, "loss": 2.1572, "step": 434 }, { "epoch": 0.04, "learning_rate": 0.001999138625174465, "loss": 2.1172, "step": 435 }, { "epoch": 0.04, "learning_rate": 0.00199912532422477, "loss": 2.2285, "step": 436 }, { "epoch": 0.04, "learning_rate": 0.001999111921411764, "loss": 2.1768, "step": 437 }, { "epoch": 0.04, "learning_rate": 0.0019990984167368143, "loss": 2.1572, "step": 438 }, { "epoch": 0.04, "learning_rate": 0.001999084810201297, "loss": 2.1455, "step": 439 }, { "epoch": 0.04, "learning_rate": 0.0019990711018065994, "loss": 2.0596, "step": 440 }, { "epoch": 0.04, "learning_rate": 0.0019990572915541192, "loss": 2.2207, "step": 441 }, { "epoch": 0.04, "learning_rate": 0.001999043379445264, "loss": 2.1895, "step": 442 }, { "epoch": 0.04, "learning_rate": 0.001999029365481453, "loss": 2.1367, "step": 443 }, { "epoch": 0.04, "learning_rate": 0.0019990152496641143, "loss": 2.1367, "step": 444 }, { "epoch": 0.04, "learning_rate": 0.001999001031994687, "loss": 2.0938, "step": 445 }, { "epoch": 0.04, "learning_rate": 0.0019989867124746213, "loss": 2.0771, "step": 446 }, { "epoch": 0.04, "learning_rate": 0.0019989722911053767, "loss": 2.207, "step": 447 }, { "epoch": 0.04, "learning_rate": 0.001998957767888423, "loss": 2.0723, "step": 448 }, { "epoch": 0.04, "learning_rate": 0.0019989431428252416, "loss": 2.1387, "step": 449 }, { "epoch": 0.04, "learning_rate": 0.0019989284159173232, "loss": 2.498, "step": 450 }, { "epoch": 0.04, "learning_rate": 0.001998913587166169, "loss": 2.1934, "step": 451 }, { "epoch": 0.04, "learning_rate": 0.0019988986565732916, "loss": 2.0605, "step": 452 }, { "epoch": 0.04, "learning_rate": 0.001998883624140213, "loss": 2.2764, "step": 453 }, { "epoch": 0.04, "learning_rate": 0.0019988684898684652, "loss": 2.1621, "step": 454 }, { "epoch": 0.04, "learning_rate": 0.0019988532537595916, "loss": 2.1914, "step": 455 }, { "epoch": 0.04, "learning_rate": 0.0019988379158151453, "loss": 2.1992, "step": 456 }, { "epoch": 0.05, "learning_rate": 0.0019988224760366905, "loss": 2.0957, "step": 457 }, { "epoch": 0.05, "learning_rate": 0.001998806934425801, "loss": 2.082, "step": 458 }, { "epoch": 0.05, "learning_rate": 0.0019987912909840617, "loss": 2.0879, "step": 459 }, { "epoch": 0.05, "learning_rate": 0.001998775545713067, "loss": 2.1562, "step": 460 }, { "epoch": 0.05, "learning_rate": 0.001998759698614422, "loss": 2.0635, "step": 461 }, { "epoch": 0.05, "learning_rate": 0.001998743749689743, "loss": 2.1543, "step": 462 }, { "epoch": 0.05, "learning_rate": 0.001998727698940656, "loss": 2.2021, "step": 463 }, { "epoch": 0.05, "learning_rate": 0.001998711546368797, "loss": 2.1035, "step": 464 }, { "epoch": 0.05, "learning_rate": 0.001998695291975812, "loss": 2.1973, "step": 465 }, { "epoch": 0.05, "learning_rate": 0.0019986789357633603, "loss": 2.1123, "step": 466 }, { "epoch": 0.05, "learning_rate": 0.0019986624777331074, "loss": 2.0273, "step": 467 }, { "epoch": 0.05, "learning_rate": 0.0019986459178867326, "loss": 2.1172, "step": 468 }, { "epoch": 0.05, "learning_rate": 0.0019986292562259233, "loss": 2.0908, "step": 469 }, { "epoch": 0.05, "learning_rate": 0.001998612492752379, "loss": 2.0537, "step": 470 }, { "epoch": 0.05, "learning_rate": 0.0019985956274678083, "loss": 2.0039, "step": 471 }, { "epoch": 0.05, "learning_rate": 0.001998578660373931, "loss": 2.1533, "step": 472 }, { "epoch": 0.05, "learning_rate": 0.0019985615914724763, "loss": 2.1182, "step": 473 }, { "epoch": 0.05, "learning_rate": 0.001998544420765185, "loss": 2.0742, "step": 474 }, { "epoch": 0.05, "learning_rate": 0.0019985271482538076, "loss": 2.0469, "step": 475 }, { "epoch": 0.05, "learning_rate": 0.0019985097739401046, "loss": 2.0742, "step": 476 }, { "epoch": 0.05, "learning_rate": 0.001998492297825848, "loss": 2.0479, "step": 477 }, { "epoch": 0.05, "learning_rate": 0.001998474719912819, "loss": 2.1484, "step": 478 }, { "epoch": 0.05, "learning_rate": 0.00199845704020281, "loss": 2.1562, "step": 479 }, { "epoch": 0.05, "learning_rate": 0.001998439258697624, "loss": 2.0713, "step": 480 }, { "epoch": 0.05, "learning_rate": 0.001998421375399073, "loss": 2.1367, "step": 481 }, { "epoch": 0.05, "learning_rate": 0.0019984033903089806, "loss": 2.1523, "step": 482 }, { "epoch": 0.05, "learning_rate": 0.0019983853034291803, "loss": 2.1094, "step": 483 }, { "epoch": 0.05, "learning_rate": 0.0019983671147615163, "loss": 2.0928, "step": 484 }, { "epoch": 0.05, "learning_rate": 0.001998348824307843, "loss": 2.0527, "step": 485 }, { "epoch": 0.05, "learning_rate": 0.0019983304320700247, "loss": 2.1162, "step": 486 }, { "epoch": 0.05, "learning_rate": 0.0019983119380499372, "loss": 2.1504, "step": 487 }, { "epoch": 0.05, "learning_rate": 0.0019982933422494657, "loss": 2.1572, "step": 488 }, { "epoch": 0.05, "learning_rate": 0.0019982746446705058, "loss": 2.127, "step": 489 }, { "epoch": 0.05, "learning_rate": 0.0019982558453149643, "loss": 2.1436, "step": 490 }, { "epoch": 0.05, "learning_rate": 0.0019982369441847573, "loss": 2.1035, "step": 491 }, { "epoch": 0.05, "learning_rate": 0.0019982179412818125, "loss": 2.0889, "step": 492 }, { "epoch": 0.05, "learning_rate": 0.0019981988366080668, "loss": 2.0518, "step": 493 }, { "epoch": 0.05, "learning_rate": 0.0019981796301654677, "loss": 2.0186, "step": 494 }, { "epoch": 0.05, "learning_rate": 0.001998160321955974, "loss": 2.165, "step": 495 }, { "epoch": 0.05, "learning_rate": 0.0019981409119815543, "loss": 2.0947, "step": 496 }, { "epoch": 0.05, "learning_rate": 0.0019981214002441866, "loss": 2.0332, "step": 497 }, { "epoch": 0.05, "learning_rate": 0.0019981017867458607, "loss": 2.1875, "step": 498 }, { "epoch": 0.05, "learning_rate": 0.0019980820714885763, "loss": 2.1514, "step": 499 }, { "epoch": 0.05, "learning_rate": 0.0019980622544743436, "loss": 2.1025, "step": 500 }, { "epoch": 0.05, "learning_rate": 0.001998042335705182, "loss": 2.0352, "step": 501 }, { "epoch": 0.05, "learning_rate": 0.001998022315183124, "loss": 2.0801, "step": 502 }, { "epoch": 0.05, "learning_rate": 0.001998002192910209, "loss": 2.0596, "step": 503 }, { "epoch": 0.05, "learning_rate": 0.0019979819688884896, "loss": 2.1387, "step": 504 }, { "epoch": 0.05, "learning_rate": 0.0019979616431200276, "loss": 2.0137, "step": 505 }, { "epoch": 0.05, "learning_rate": 0.0019979412156068952, "loss": 2.084, "step": 506 }, { "epoch": 0.05, "learning_rate": 0.0019979206863511745, "loss": 2.0557, "step": 507 }, { "epoch": 0.05, "learning_rate": 0.0019979000553549584, "loss": 2.0928, "step": 508 }, { "epoch": 0.05, "learning_rate": 0.001997879322620351, "loss": 2.0654, "step": 509 }, { "epoch": 0.05, "learning_rate": 0.0019978584881494665, "loss": 2.0938, "step": 510 }, { "epoch": 0.05, "learning_rate": 0.001997837551944428, "loss": 2.0303, "step": 511 }, { "epoch": 0.05, "learning_rate": 0.00199781651400737, "loss": 2.0967, "step": 512 }, { "epoch": 0.05, "learning_rate": 0.0019977953743404377, "loss": 2.1338, "step": 513 }, { "epoch": 0.05, "learning_rate": 0.0019977741329457867, "loss": 2.1025, "step": 514 }, { "epoch": 0.05, "learning_rate": 0.001997752789825582, "loss": 2.0957, "step": 515 }, { "epoch": 0.05, "learning_rate": 0.001997731344982, "loss": 2.0547, "step": 516 }, { "epoch": 0.05, "learning_rate": 0.0019977097984172267, "loss": 2.0557, "step": 517 }, { "epoch": 0.05, "learning_rate": 0.001997688150133459, "loss": 2.0205, "step": 518 }, { "epoch": 0.05, "learning_rate": 0.001997666400132904, "loss": 2.1201, "step": 519 }, { "epoch": 0.05, "learning_rate": 0.0019976445484177797, "loss": 2.168, "step": 520 }, { "epoch": 0.05, "learning_rate": 0.001997622594990313, "loss": 2.1367, "step": 521 }, { "epoch": 0.05, "learning_rate": 0.001997600539852743, "loss": 2.1123, "step": 522 }, { "epoch": 0.05, "learning_rate": 0.001997578383007317, "loss": 2.0156, "step": 523 }, { "epoch": 0.05, "learning_rate": 0.0019975561244562953, "loss": 2.1143, "step": 524 }, { "epoch": 0.05, "learning_rate": 0.0019975337642019466, "loss": 2.0703, "step": 525 }, { "epoch": 0.05, "learning_rate": 0.0019975113022465507, "loss": 2.0137, "step": 526 }, { "epoch": 0.05, "learning_rate": 0.0019974887385923973, "loss": 2.0732, "step": 527 }, { "epoch": 0.05, "learning_rate": 0.0019974660732417875, "loss": 2.0518, "step": 528 }, { "epoch": 0.05, "learning_rate": 0.0019974433061970315, "loss": 2.1387, "step": 529 }, { "epoch": 0.05, "learning_rate": 0.001997420437460451, "loss": 2.1123, "step": 530 }, { "epoch": 0.05, "learning_rate": 0.0019973974670343766, "loss": 2.0957, "step": 531 }, { "epoch": 0.05, "learning_rate": 0.001997374394921151, "loss": 2.0615, "step": 532 }, { "epoch": 0.05, "learning_rate": 0.0019973512211231264, "loss": 2.0908, "step": 533 }, { "epoch": 0.05, "learning_rate": 0.001997327945642665, "loss": 2.082, "step": 534 }, { "epoch": 0.05, "learning_rate": 0.0019973045684821404, "loss": 2.1494, "step": 535 }, { "epoch": 0.05, "learning_rate": 0.0019972810896439356, "loss": 2.0508, "step": 536 }, { "epoch": 0.05, "learning_rate": 0.001997257509130444, "loss": 2.0459, "step": 537 }, { "epoch": 0.05, "learning_rate": 0.0019972338269440702, "loss": 2.0967, "step": 538 }, { "epoch": 0.05, "learning_rate": 0.001997210043087229, "loss": 2.0137, "step": 539 }, { "epoch": 0.05, "learning_rate": 0.0019971861575623444, "loss": 2.0869, "step": 540 }, { "epoch": 0.05, "learning_rate": 0.0019971621703718514, "loss": 2.084, "step": 541 }, { "epoch": 0.05, "learning_rate": 0.0019971380815181966, "loss": 2.0654, "step": 542 }, { "epoch": 0.05, "learning_rate": 0.001997113891003835, "loss": 1.998, "step": 543 }, { "epoch": 0.05, "learning_rate": 0.001997089598831234, "loss": 2.0674, "step": 544 }, { "epoch": 0.05, "learning_rate": 0.0019970652050028685, "loss": 2.0186, "step": 545 }, { "epoch": 0.05, "learning_rate": 0.001997040709521227, "loss": 2.1221, "step": 546 }, { "epoch": 0.05, "learning_rate": 0.0019970161123888065, "loss": 2.165, "step": 547 }, { "epoch": 0.05, "learning_rate": 0.001996991413608115, "loss": 2.0459, "step": 548 }, { "epoch": 0.05, "learning_rate": 0.0019969666131816697, "loss": 2.0635, "step": 549 }, { "epoch": 0.05, "learning_rate": 0.0019969417111119995, "loss": 2.0566, "step": 550 }, { "epoch": 0.05, "learning_rate": 0.001996916707401644, "loss": 2.1035, "step": 551 }, { "epoch": 0.05, "learning_rate": 0.001996891602053151, "loss": 2.1104, "step": 552 }, { "epoch": 0.05, "learning_rate": 0.001996866395069081, "loss": 2.042, "step": 553 }, { "epoch": 0.05, "learning_rate": 0.0019968410864520036, "loss": 2.0176, "step": 554 }, { "epoch": 0.05, "learning_rate": 0.0019968156762044994, "loss": 2.1416, "step": 555 }, { "epoch": 0.05, "learning_rate": 0.001996790164329159, "loss": 1.9971, "step": 556 }, { "epoch": 0.05, "learning_rate": 0.0019967645508285827, "loss": 2.0967, "step": 557 }, { "epoch": 0.06, "learning_rate": 0.001996738835705383, "loss": 2.124, "step": 558 }, { "epoch": 0.06, "learning_rate": 0.00199671301896218, "loss": 2.1016, "step": 559 }, { "epoch": 0.06, "learning_rate": 0.001996687100601608, "loss": 2.0645, "step": 560 }, { "epoch": 0.06, "learning_rate": 0.001996661080626307, "loss": 2.1006, "step": 561 }, { "epoch": 0.06, "learning_rate": 0.0019966349590389318, "loss": 2.0205, "step": 562 }, { "epoch": 0.06, "learning_rate": 0.0019966087358421446, "loss": 2.0254, "step": 563 }, { "epoch": 0.06, "learning_rate": 0.0019965824110386187, "loss": 2.0527, "step": 564 }, { "epoch": 0.06, "learning_rate": 0.0019965559846310387, "loss": 2.0801, "step": 565 }, { "epoch": 0.06, "learning_rate": 0.0019965294566220986, "loss": 2.1094, "step": 566 }, { "epoch": 0.06, "learning_rate": 0.001996502827014503, "loss": 2.0645, "step": 567 }, { "epoch": 0.06, "learning_rate": 0.001996476095810967, "loss": 2.1143, "step": 568 }, { "epoch": 0.06, "learning_rate": 0.001996449263014215, "loss": 2.0098, "step": 569 }, { "epoch": 0.06, "learning_rate": 0.0019964223286269836, "loss": 2.1699, "step": 570 }, { "epoch": 0.06, "learning_rate": 0.0019963952926520183, "loss": 2.0635, "step": 571 }, { "epoch": 0.06, "learning_rate": 0.001996368155092076, "loss": 2.0537, "step": 572 }, { "epoch": 0.06, "learning_rate": 0.001996340915949923, "loss": 2.1309, "step": 573 }, { "epoch": 0.06, "learning_rate": 0.001996313575228337, "loss": 2.041, "step": 574 }, { "epoch": 0.06, "learning_rate": 0.001996286132930105, "loss": 2.1357, "step": 575 }, { "epoch": 0.06, "learning_rate": 0.001996258589058025, "loss": 1.9932, "step": 576 }, { "epoch": 0.06, "learning_rate": 0.0019962309436149044, "loss": 1.9775, "step": 577 }, { "epoch": 0.06, "learning_rate": 0.0019962031966035626, "loss": 2.0068, "step": 578 }, { "epoch": 0.06, "learning_rate": 0.0019961753480268285, "loss": 2.083, "step": 579 }, { "epoch": 0.06, "learning_rate": 0.001996147397887541, "loss": 2.0391, "step": 580 }, { "epoch": 0.06, "learning_rate": 0.0019961193461885494, "loss": 2.0488, "step": 581 }, { "epoch": 0.06, "learning_rate": 0.0019960911929327143, "loss": 2.1201, "step": 582 }, { "epoch": 0.06, "learning_rate": 0.0019960629381229056, "loss": 2.0508, "step": 583 }, { "epoch": 0.06, "learning_rate": 0.001996034581762004, "loss": 2.1035, "step": 584 }, { "epoch": 0.06, "learning_rate": 0.0019960061238529004, "loss": 2.0781, "step": 585 }, { "epoch": 0.06, "learning_rate": 0.001995977564398496, "loss": 2.0469, "step": 586 }, { "epoch": 0.06, "learning_rate": 0.0019959489034017037, "loss": 2.084, "step": 587 }, { "epoch": 0.06, "learning_rate": 0.001995920140865444, "loss": 2.0303, "step": 588 }, { "epoch": 0.06, "learning_rate": 0.00199589127679265, "loss": 2.0059, "step": 589 }, { "epoch": 0.06, "learning_rate": 0.0019958623111862644, "loss": 2.0137, "step": 590 }, { "epoch": 0.06, "learning_rate": 0.0019958332440492407, "loss": 1.96, "step": 591 }, { "epoch": 0.06, "learning_rate": 0.0019958040753845414, "loss": 2.0752, "step": 592 }, { "epoch": 0.06, "learning_rate": 0.0019957748051951415, "loss": 2.0449, "step": 593 }, { "epoch": 0.06, "learning_rate": 0.0019957454334840244, "loss": 2.0693, "step": 594 }, { "epoch": 0.06, "learning_rate": 0.0019957159602541847, "loss": 2.0762, "step": 595 }, { "epoch": 0.06, "learning_rate": 0.0019956863855086276, "loss": 2.0869, "step": 596 }, { "epoch": 0.06, "learning_rate": 0.001995656709250368, "loss": 2.0518, "step": 597 }, { "epoch": 0.06, "learning_rate": 0.0019956269314824315, "loss": 2.0127, "step": 598 }, { "epoch": 0.06, "learning_rate": 0.0019955970522078544, "loss": 2.0918, "step": 599 }, { "epoch": 0.06, "learning_rate": 0.0019955670714296822, "loss": 1.9941, "step": 600 }, { "epoch": 0.06, "learning_rate": 0.001995536989150972, "loss": 2.0518, "step": 601 }, { "epoch": 0.06, "learning_rate": 0.001995506805374791, "loss": 1.9932, "step": 602 }, { "epoch": 0.06, "learning_rate": 0.001995476520104216, "loss": 2.0771, "step": 603 }, { "epoch": 0.06, "learning_rate": 0.001995446133342335, "loss": 2.0801, "step": 604 }, { "epoch": 0.06, "learning_rate": 0.0019954156450922456, "loss": 2.0664, "step": 605 }, { "epoch": 0.06, "learning_rate": 0.001995385055357057, "loss": 2.041, "step": 606 }, { "epoch": 0.06, "learning_rate": 0.001995354364139887, "loss": 2.0439, "step": 607 }, { "epoch": 0.06, "learning_rate": 0.0019953235714438656, "loss": 2.0293, "step": 608 }, { "epoch": 0.06, "learning_rate": 0.001995292677272131, "loss": 2.0557, "step": 609 }, { "epoch": 0.06, "learning_rate": 0.0019952616816278335, "loss": 2.0039, "step": 610 }, { "epoch": 0.06, "learning_rate": 0.0019952305845141335, "loss": 2.0029, "step": 611 }, { "epoch": 0.06, "learning_rate": 0.001995199385934201, "loss": 2.0137, "step": 612 }, { "epoch": 0.06, "learning_rate": 0.0019951680858912167, "loss": 1.9688, "step": 613 }, { "epoch": 0.06, "learning_rate": 0.0019951366843883726, "loss": 2.085, "step": 614 }, { "epoch": 0.06, "learning_rate": 0.001995105181428869, "loss": 2.0547, "step": 615 }, { "epoch": 0.06, "learning_rate": 0.001995073577015918, "loss": 2.0225, "step": 616 }, { "epoch": 0.06, "learning_rate": 0.0019950418711527423, "loss": 2.0781, "step": 617 }, { "epoch": 0.06, "learning_rate": 0.001995010063842574, "loss": 1.9971, "step": 618 }, { "epoch": 0.06, "learning_rate": 0.001994978155088656, "loss": 2.0488, "step": 619 }, { "epoch": 0.06, "learning_rate": 0.0019949461448942416, "loss": 2.1377, "step": 620 }, { "epoch": 0.06, "learning_rate": 0.001994914033262594, "loss": 2.0527, "step": 621 }, { "epoch": 0.06, "learning_rate": 0.001994881820196987, "loss": 1.9277, "step": 622 }, { "epoch": 0.06, "learning_rate": 0.001994849505700705, "loss": 2.0342, "step": 623 }, { "epoch": 0.06, "learning_rate": 0.001994817089777043, "loss": 2.0508, "step": 624 }, { "epoch": 0.06, "learning_rate": 0.0019947845724293057, "loss": 2.04, "step": 625 }, { "epoch": 0.06, "learning_rate": 0.0019947519536608073, "loss": 2.0674, "step": 626 }, { "epoch": 0.06, "learning_rate": 0.0019947192334748747, "loss": 2.0264, "step": 627 }, { "epoch": 0.06, "learning_rate": 0.001994686411874843, "loss": 1.9863, "step": 628 }, { "epoch": 0.06, "learning_rate": 0.0019946534888640585, "loss": 2.0498, "step": 629 }, { "epoch": 0.06, "learning_rate": 0.0019946204644458784, "loss": 2.0137, "step": 630 }, { "epoch": 0.06, "learning_rate": 0.0019945873386236694, "loss": 2.0049, "step": 631 }, { "epoch": 0.06, "learning_rate": 0.001994554111400808, "loss": 2.0391, "step": 632 }, { "epoch": 0.06, "learning_rate": 0.0019945207827806828, "loss": 2.0273, "step": 633 }, { "epoch": 0.06, "learning_rate": 0.0019944873527666905, "loss": 2.0654, "step": 634 }, { "epoch": 0.06, "learning_rate": 0.001994453821362241, "loss": 1.9814, "step": 635 }, { "epoch": 0.06, "learning_rate": 0.0019944201885707517, "loss": 1.9912, "step": 636 }, { "epoch": 0.06, "learning_rate": 0.001994386454395652, "loss": 1.9658, "step": 637 }, { "epoch": 0.06, "learning_rate": 0.0019943526188403813, "loss": 2.0254, "step": 638 }, { "epoch": 0.06, "learning_rate": 0.001994318681908389, "loss": 2.0566, "step": 639 }, { "epoch": 0.06, "learning_rate": 0.0019942846436031346, "loss": 1.9961, "step": 640 }, { "epoch": 0.06, "learning_rate": 0.0019942505039280893, "loss": 2.0488, "step": 641 }, { "epoch": 0.06, "learning_rate": 0.0019942162628867333, "loss": 2.042, "step": 642 }, { "epoch": 0.06, "learning_rate": 0.0019941819204825576, "loss": 2.0557, "step": 643 }, { "epoch": 0.06, "learning_rate": 0.0019941474767190635, "loss": 1.9404, "step": 644 }, { "epoch": 0.06, "learning_rate": 0.0019941129315997623, "loss": 2.0088, "step": 645 }, { "epoch": 0.06, "learning_rate": 0.001994078285128176, "loss": 1.9873, "step": 646 }, { "epoch": 0.06, "learning_rate": 0.001994043537307838, "loss": 2.041, "step": 647 }, { "epoch": 0.06, "learning_rate": 0.001994008688142289, "loss": 1.9648, "step": 648 }, { "epoch": 0.06, "learning_rate": 0.0019939737376350837, "loss": 2.0312, "step": 649 }, { "epoch": 0.06, "learning_rate": 0.0019939386857897846, "loss": 2.0068, "step": 650 }, { "epoch": 0.06, "learning_rate": 0.0019939035326099657, "loss": 2.1006, "step": 651 }, { "epoch": 0.06, "learning_rate": 0.00199386827809921, "loss": 2.0283, "step": 652 }, { "epoch": 0.06, "learning_rate": 0.001993832922261113, "loss": 1.9648, "step": 653 }, { "epoch": 0.06, "learning_rate": 0.0019937974650992784, "loss": 1.9902, "step": 654 }, { "epoch": 0.06, "learning_rate": 0.001993761906617322, "loss": 2.0088, "step": 655 }, { "epoch": 0.06, "learning_rate": 0.001993726246818868, "loss": 1.9453, "step": 656 }, { "epoch": 0.06, "learning_rate": 0.0019936904857075534, "loss": 2.0098, "step": 657 }, { "epoch": 0.06, "learning_rate": 0.0019936546232870228, "loss": 1.9785, "step": 658 }, { "epoch": 0.06, "learning_rate": 0.001993618659560933, "loss": 2.0352, "step": 659 }, { "epoch": 0.07, "learning_rate": 0.001993582594532951, "loss": 2.1367, "step": 660 }, { "epoch": 0.07, "learning_rate": 0.001993546428206753, "loss": 2.0391, "step": 661 }, { "epoch": 0.07, "learning_rate": 0.0019935101605860267, "loss": 2.0254, "step": 662 }, { "epoch": 0.07, "learning_rate": 0.0019934737916744694, "loss": 2.0576, "step": 663 }, { "epoch": 0.07, "learning_rate": 0.0019934373214757894, "loss": 2.0576, "step": 664 }, { "epoch": 0.07, "learning_rate": 0.0019934007499937045, "loss": 2.0078, "step": 665 }, { "epoch": 0.07, "learning_rate": 0.001993364077231943, "loss": 2.0439, "step": 666 }, { "epoch": 0.07, "learning_rate": 0.001993327303194245, "loss": 2.1016, "step": 667 }, { "epoch": 0.07, "learning_rate": 0.0019932904278843586, "loss": 2.0264, "step": 668 }, { "epoch": 0.07, "learning_rate": 0.0019932534513060433, "loss": 2.0283, "step": 669 }, { "epoch": 0.07, "learning_rate": 0.00199321637346307, "loss": 2.0029, "step": 670 }, { "epoch": 0.07, "learning_rate": 0.0019931791943592174, "loss": 2.04, "step": 671 }, { "epoch": 0.07, "learning_rate": 0.0019931419139982767, "loss": 2.0205, "step": 672 }, { "epoch": 0.07, "learning_rate": 0.0019931045323840495, "loss": 2.1992, "step": 673 }, { "epoch": 0.07, "learning_rate": 0.001993067049520346, "loss": 2.0498, "step": 674 }, { "epoch": 0.07, "learning_rate": 0.001993029465410988, "loss": 2.0645, "step": 675 }, { "epoch": 0.07, "learning_rate": 0.0019929917800598066, "loss": 2.0439, "step": 676 }, { "epoch": 0.07, "learning_rate": 0.001992953993470645, "loss": 1.9814, "step": 677 }, { "epoch": 0.07, "learning_rate": 0.0019929161056473554, "loss": 2.0615, "step": 678 }, { "epoch": 0.07, "learning_rate": 0.0019928781165937997, "loss": 2.0957, "step": 679 }, { "epoch": 0.07, "learning_rate": 0.001992840026313852, "loss": 2.0703, "step": 680 }, { "epoch": 0.07, "learning_rate": 0.0019928018348113956, "loss": 2.0947, "step": 681 }, { "epoch": 0.07, "learning_rate": 0.0019927635420903234, "loss": 2.0342, "step": 682 }, { "epoch": 0.07, "learning_rate": 0.00199272514815454, "loss": 1.9355, "step": 683 }, { "epoch": 0.07, "learning_rate": 0.0019926866530079602, "loss": 2.0029, "step": 684 }, { "epoch": 0.07, "learning_rate": 0.0019926480566545075, "loss": 2.04, "step": 685 }, { "epoch": 0.07, "learning_rate": 0.001992609359098118, "loss": 2.0908, "step": 686 }, { "epoch": 0.07, "learning_rate": 0.0019925705603427366, "loss": 2.1064, "step": 687 }, { "epoch": 0.07, "learning_rate": 0.0019925316603923186, "loss": 2.0654, "step": 688 }, { "epoch": 0.07, "learning_rate": 0.0019924926592508304, "loss": 1.9639, "step": 689 }, { "epoch": 0.07, "learning_rate": 0.0019924535569222483, "loss": 2.0791, "step": 690 }, { "epoch": 0.07, "learning_rate": 0.0019924143534105587, "loss": 2.0137, "step": 691 }, { "epoch": 0.07, "learning_rate": 0.0019923750487197585, "loss": 1.9854, "step": 692 }, { "epoch": 0.07, "learning_rate": 0.001992335642853855, "loss": 2.0371, "step": 693 }, { "epoch": 0.07, "learning_rate": 0.0019922961358168653, "loss": 1.9346, "step": 694 }, { "epoch": 0.07, "learning_rate": 0.0019922565276128176, "loss": 1.9336, "step": 695 }, { "epoch": 0.07, "learning_rate": 0.0019922168182457504, "loss": 2.0117, "step": 696 }, { "epoch": 0.07, "learning_rate": 0.0019921770077197115, "loss": 1.9854, "step": 697 }, { "epoch": 0.07, "learning_rate": 0.00199213709603876, "loss": 1.9512, "step": 698 }, { "epoch": 0.07, "learning_rate": 0.001992097083206965, "loss": 2.0918, "step": 699 }, { "epoch": 0.07, "learning_rate": 0.001992056969228406, "loss": 2.0391, "step": 700 }, { "epoch": 0.07, "learning_rate": 0.0019920167541071726, "loss": 2.0703, "step": 701 }, { "epoch": 0.07, "learning_rate": 0.0019919764378473655, "loss": 2.0146, "step": 702 }, { "epoch": 0.07, "learning_rate": 0.0019919360204530934, "loss": 2.0293, "step": 703 }, { "epoch": 0.07, "learning_rate": 0.0019918955019284784, "loss": 2.0205, "step": 704 }, { "epoch": 0.07, "learning_rate": 0.0019918548822776506, "loss": 2.0479, "step": 705 }, { "epoch": 0.07, "learning_rate": 0.001991814161504752, "loss": 2.0361, "step": 706 }, { "epoch": 0.07, "learning_rate": 0.001991773339613934, "loss": 2.0127, "step": 707 }, { "epoch": 0.07, "learning_rate": 0.001991732416609358, "loss": 1.9648, "step": 708 }, { "epoch": 0.07, "learning_rate": 0.001991691392495197, "loss": 1.9736, "step": 709 }, { "epoch": 0.07, "learning_rate": 0.001991650267275633, "loss": 2.002, "step": 710 }, { "epoch": 0.07, "learning_rate": 0.0019916090409548585, "loss": 1.9727, "step": 711 }, { "epoch": 0.07, "learning_rate": 0.001991567713537077, "loss": 2.0039, "step": 712 }, { "epoch": 0.07, "learning_rate": 0.0019915262850265024, "loss": 2.0176, "step": 713 }, { "epoch": 0.07, "learning_rate": 0.001991484755427358, "loss": 1.9043, "step": 714 }, { "epoch": 0.07, "learning_rate": 0.0019914431247438774, "loss": 2.0322, "step": 715 }, { "epoch": 0.07, "learning_rate": 0.0019914013929803053, "loss": 1.9346, "step": 716 }, { "epoch": 0.07, "learning_rate": 0.001991359560140897, "loss": 2.0283, "step": 717 }, { "epoch": 0.07, "learning_rate": 0.0019913176262299163, "loss": 2.0146, "step": 718 }, { "epoch": 0.07, "learning_rate": 0.0019912755912516395, "loss": 2.0215, "step": 719 }, { "epoch": 0.07, "learning_rate": 0.0019912334552103514, "loss": 2.0264, "step": 720 }, { "epoch": 0.07, "learning_rate": 0.0019911912181103486, "loss": 2.0098, "step": 721 }, { "epoch": 0.07, "learning_rate": 0.0019911488799559366, "loss": 1.8994, "step": 722 }, { "epoch": 0.07, "learning_rate": 0.0019911064407514326, "loss": 2.0801, "step": 723 }, { "epoch": 0.07, "learning_rate": 0.0019910639005011627, "loss": 2.0273, "step": 724 }, { "epoch": 0.07, "learning_rate": 0.001991021259209464, "loss": 2.1045, "step": 725 }, { "epoch": 0.07, "learning_rate": 0.001990978516880684, "loss": 1.9961, "step": 726 }, { "epoch": 0.07, "learning_rate": 0.001990935673519181, "loss": 1.9434, "step": 727 }, { "epoch": 0.07, "learning_rate": 0.001990892729129322, "loss": 1.9834, "step": 728 }, { "epoch": 0.07, "learning_rate": 0.0019908496837154864, "loss": 2.0664, "step": 729 }, { "epoch": 0.07, "learning_rate": 0.001990806537282062, "loss": 1.8936, "step": 730 }, { "epoch": 0.07, "learning_rate": 0.0019907632898334476, "loss": 1.8955, "step": 731 }, { "epoch": 0.07, "learning_rate": 0.0019907199413740527, "loss": 2.042, "step": 732 }, { "epoch": 0.07, "learning_rate": 0.0019906764919082973, "loss": 1.9541, "step": 733 }, { "epoch": 0.07, "learning_rate": 0.00199063294144061, "loss": 1.8828, "step": 734 }, { "epoch": 0.07, "learning_rate": 0.0019905892899754318, "loss": 2.0273, "step": 735 }, { "epoch": 0.07, "learning_rate": 0.0019905455375172124, "loss": 1.8936, "step": 736 }, { "epoch": 0.07, "learning_rate": 0.0019905016840704127, "loss": 1.9961, "step": 737 }, { "epoch": 0.07, "learning_rate": 0.0019904577296395044, "loss": 2.0254, "step": 738 }, { "epoch": 0.07, "learning_rate": 0.0019904136742289677, "loss": 1.8975, "step": 739 }, { "epoch": 0.07, "learning_rate": 0.001990369517843295, "loss": 2.0195, "step": 740 }, { "epoch": 0.07, "learning_rate": 0.0019903252604869873, "loss": 2.0264, "step": 741 }, { "epoch": 0.07, "learning_rate": 0.0019902809021645576, "loss": 1.9658, "step": 742 }, { "epoch": 0.07, "learning_rate": 0.001990236442880528, "loss": 1.9883, "step": 743 }, { "epoch": 0.07, "learning_rate": 0.001990191882639431, "loss": 2.0381, "step": 744 }, { "epoch": 0.07, "learning_rate": 0.00199014722144581, "loss": 2.0039, "step": 745 }, { "epoch": 0.07, "learning_rate": 0.001990102459304218, "loss": 2.0508, "step": 746 }, { "epoch": 0.07, "learning_rate": 0.0019900575962192185, "loss": 2.0195, "step": 747 }, { "epoch": 0.07, "learning_rate": 0.001990012632195386, "loss": 1.9922, "step": 748 }, { "epoch": 0.07, "learning_rate": 0.0019899675672373045, "loss": 2.1113, "step": 749 }, { "epoch": 0.07, "learning_rate": 0.0019899224013495685, "loss": 1.9902, "step": 750 }, { "epoch": 0.07, "learning_rate": 0.001989877134536782, "loss": 2.0273, "step": 751 }, { "epoch": 0.07, "learning_rate": 0.001989831766803561, "loss": 2.0654, "step": 752 }, { "epoch": 0.07, "learning_rate": 0.0019897862981545305, "loss": 1.9941, "step": 753 }, { "epoch": 0.07, "learning_rate": 0.001989740728594326, "loss": 2.0781, "step": 754 }, { "epoch": 0.07, "learning_rate": 0.0019896950581275935, "loss": 2.0254, "step": 755 }, { "epoch": 0.07, "learning_rate": 0.00198964928675899, "loss": 1.9658, "step": 756 }, { "epoch": 0.07, "learning_rate": 0.0019896034144931806, "loss": 2.0234, "step": 757 }, { "epoch": 0.07, "learning_rate": 0.001989557441334843, "loss": 2.0586, "step": 758 }, { "epoch": 0.07, "learning_rate": 0.0019895113672886646, "loss": 1.9453, "step": 759 }, { "epoch": 0.07, "learning_rate": 0.0019894651923593417, "loss": 1.959, "step": 760 }, { "epoch": 0.08, "learning_rate": 0.0019894189165515826, "loss": 2.0352, "step": 761 }, { "epoch": 0.08, "learning_rate": 0.0019893725398701053, "loss": 1.9834, "step": 762 }, { "epoch": 0.08, "learning_rate": 0.0019893260623196378, "loss": 1.8203, "step": 763 }, { "epoch": 0.08, "learning_rate": 0.001989279483904919, "loss": 2.0801, "step": 764 }, { "epoch": 0.08, "learning_rate": 0.0019892328046306968, "loss": 2.1152, "step": 765 }, { "epoch": 0.08, "learning_rate": 0.001989186024501731, "loss": 2.0371, "step": 766 }, { "epoch": 0.08, "learning_rate": 0.001989139143522791, "loss": 2.0049, "step": 767 }, { "epoch": 0.08, "learning_rate": 0.001989092161698656, "loss": 2.0059, "step": 768 }, { "epoch": 0.08, "learning_rate": 0.0019890450790341157, "loss": 1.9287, "step": 769 }, { "epoch": 0.08, "learning_rate": 0.001988997895533971, "loss": 1.9951, "step": 770 }, { "epoch": 0.08, "learning_rate": 0.0019889506112030324, "loss": 1.9209, "step": 771 }, { "epoch": 0.08, "learning_rate": 0.00198890322604612, "loss": 2.0254, "step": 772 }, { "epoch": 0.08, "learning_rate": 0.0019888557400680653, "loss": 2.0752, "step": 773 }, { "epoch": 0.08, "learning_rate": 0.0019888081532737096, "loss": 1.9775, "step": 774 }, { "epoch": 0.08, "learning_rate": 0.001988760465667904, "loss": 1.9912, "step": 775 }, { "epoch": 0.08, "learning_rate": 0.001988712677255511, "loss": 2.0098, "step": 776 }, { "epoch": 0.08, "learning_rate": 0.0019886647880414023, "loss": 2.0312, "step": 777 }, { "epoch": 0.08, "learning_rate": 0.0019886167980304606, "loss": 1.9082, "step": 778 }, { "epoch": 0.08, "learning_rate": 0.001988568707227579, "loss": 1.9756, "step": 779 }, { "epoch": 0.08, "learning_rate": 0.0019885205156376594, "loss": 1.8682, "step": 780 }, { "epoch": 0.08, "learning_rate": 0.001988472223265616, "loss": 2.0479, "step": 781 }, { "epoch": 0.08, "learning_rate": 0.001988423830116372, "loss": 2.0264, "step": 782 }, { "epoch": 0.08, "learning_rate": 0.0019883753361948607, "loss": 2.042, "step": 783 }, { "epoch": 0.08, "learning_rate": 0.001988326741506027, "loss": 1.9355, "step": 784 }, { "epoch": 0.08, "learning_rate": 0.001988278046054825, "loss": 2.0049, "step": 785 }, { "epoch": 0.08, "learning_rate": 0.001988229249846219, "loss": 1.9404, "step": 786 }, { "epoch": 0.08, "learning_rate": 0.0019881803528851846, "loss": 1.9941, "step": 787 }, { "epoch": 0.08, "learning_rate": 0.001988131355176706, "loss": 1.9814, "step": 788 }, { "epoch": 0.08, "learning_rate": 0.001988082256725779, "loss": 1.9775, "step": 789 }, { "epoch": 0.08, "learning_rate": 0.00198803305753741, "loss": 2.0459, "step": 790 }, { "epoch": 0.08, "learning_rate": 0.0019879837576166136, "loss": 1.9736, "step": 791 }, { "epoch": 0.08, "learning_rate": 0.0019879343569684177, "loss": 2.0684, "step": 792 }, { "epoch": 0.08, "learning_rate": 0.0019878848555978574, "loss": 1.9854, "step": 793 }, { "epoch": 0.08, "learning_rate": 0.0019878352535099802, "loss": 2.001, "step": 794 }, { "epoch": 0.08, "learning_rate": 0.0019877855507098433, "loss": 1.918, "step": 795 }, { "epoch": 0.08, "learning_rate": 0.0019877357472025134, "loss": 2.0615, "step": 796 }, { "epoch": 0.08, "learning_rate": 0.0019876858429930685, "loss": 1.9717, "step": 797 }, { "epoch": 0.08, "learning_rate": 0.001987635838086597, "loss": 2.0137, "step": 798 }, { "epoch": 0.08, "learning_rate": 0.0019875857324881956, "loss": 2.082, "step": 799 }, { "epoch": 0.08, "learning_rate": 0.001987535526202974, "loss": 2.0361, "step": 800 }, { "epoch": 0.08, "learning_rate": 0.00198748521923605, "loss": 2.0059, "step": 801 }, { "epoch": 0.08, "learning_rate": 0.0019874348115925533, "loss": 2.125, "step": 802 }, { "epoch": 0.08, "learning_rate": 0.0019873843032776227, "loss": 2.0049, "step": 803 }, { "epoch": 0.08, "learning_rate": 0.0019873336942964072, "loss": 2.0684, "step": 804 }, { "epoch": 0.08, "learning_rate": 0.0019872829846540674, "loss": 2.0488, "step": 805 }, { "epoch": 0.08, "learning_rate": 0.0019872321743557722, "loss": 2.0381, "step": 806 }, { "epoch": 0.08, "learning_rate": 0.001987181263406703, "loss": 1.9414, "step": 807 }, { "epoch": 0.08, "learning_rate": 0.0019871302518120497, "loss": 1.9443, "step": 808 }, { "epoch": 0.08, "learning_rate": 0.001987079139577013, "loss": 1.9619, "step": 809 }, { "epoch": 0.08, "learning_rate": 0.0019870279267068035, "loss": 1.9834, "step": 810 }, { "epoch": 0.08, "learning_rate": 0.0019869766132066436, "loss": 2.0166, "step": 811 }, { "epoch": 0.08, "learning_rate": 0.001986925199081764, "loss": 2.1025, "step": 812 }, { "epoch": 0.08, "learning_rate": 0.0019868736843374067, "loss": 1.958, "step": 813 }, { "epoch": 0.08, "learning_rate": 0.001986822068978824, "loss": 1.9922, "step": 814 }, { "epoch": 0.08, "learning_rate": 0.0019867703530112783, "loss": 1.8418, "step": 815 }, { "epoch": 0.08, "learning_rate": 0.0019867185364400414, "loss": 1.9814, "step": 816 }, { "epoch": 0.08, "learning_rate": 0.001986666619270397, "loss": 1.9883, "step": 817 }, { "epoch": 0.08, "learning_rate": 0.0019866146015076374, "loss": 2.0322, "step": 818 }, { "epoch": 0.08, "learning_rate": 0.0019865624831570663, "loss": 1.9912, "step": 819 }, { "epoch": 0.08, "learning_rate": 0.001986510264223997, "loss": 2.0674, "step": 820 }, { "epoch": 0.08, "learning_rate": 0.0019864579447137548, "loss": 2.0312, "step": 821 }, { "epoch": 0.08, "learning_rate": 0.001986405524631672, "loss": 2.0059, "step": 822 }, { "epoch": 0.08, "learning_rate": 0.0019863530039830935, "loss": 2.0293, "step": 823 }, { "epoch": 0.08, "learning_rate": 0.0019863003827733745, "loss": 2.0088, "step": 824 }, { "epoch": 0.08, "learning_rate": 0.001986247661007879, "loss": 1.9463, "step": 825 }, { "epoch": 0.08, "learning_rate": 0.0019861948386919828, "loss": 1.9316, "step": 826 }, { "epoch": 0.08, "learning_rate": 0.0019861419158310706, "loss": 1.9824, "step": 827 }, { "epoch": 0.08, "learning_rate": 0.0019860888924305388, "loss": 2.0723, "step": 828 }, { "epoch": 0.08, "learning_rate": 0.0019860357684957928, "loss": 1.8926, "step": 829 }, { "epoch": 0.08, "learning_rate": 0.001985982544032249, "loss": 2.0547, "step": 830 }, { "epoch": 0.08, "learning_rate": 0.001985929219045333, "loss": 2.0137, "step": 831 }, { "epoch": 0.08, "learning_rate": 0.0019858757935404822, "loss": 2.0225, "step": 832 }, { "epoch": 0.08, "learning_rate": 0.0019858222675231433, "loss": 1.9824, "step": 833 }, { "epoch": 0.08, "learning_rate": 0.0019857686409987738, "loss": 2.0264, "step": 834 }, { "epoch": 0.08, "learning_rate": 0.00198571491397284, "loss": 1.9395, "step": 835 }, { "epoch": 0.08, "learning_rate": 0.0019856610864508206, "loss": 1.9941, "step": 836 }, { "epoch": 0.08, "learning_rate": 0.0019856071584382026, "loss": 1.9561, "step": 837 }, { "epoch": 0.08, "learning_rate": 0.0019855531299404847, "loss": 2.002, "step": 838 }, { "epoch": 0.08, "learning_rate": 0.001985499000963175, "loss": 2.0459, "step": 839 }, { "epoch": 0.08, "learning_rate": 0.001985444771511792, "loss": 1.9707, "step": 840 }, { "epoch": 0.08, "learning_rate": 0.001985390441591864, "loss": 2.0225, "step": 841 }, { "epoch": 0.08, "learning_rate": 0.0019853360112089312, "loss": 1.9551, "step": 842 }, { "epoch": 0.08, "learning_rate": 0.0019852814803685425, "loss": 1.9863, "step": 843 }, { "epoch": 0.08, "learning_rate": 0.0019852268490762573, "loss": 1.9834, "step": 844 }, { "epoch": 0.08, "learning_rate": 0.0019851721173376456, "loss": 2.0557, "step": 845 }, { "epoch": 0.08, "learning_rate": 0.0019851172851582865, "loss": 2.0166, "step": 846 }, { "epoch": 0.08, "learning_rate": 0.001985062352543772, "loss": 1.8398, "step": 847 }, { "epoch": 0.08, "learning_rate": 0.001985007319499701, "loss": 1.9297, "step": 848 }, { "epoch": 0.08, "learning_rate": 0.0019849521860316856, "loss": 1.9619, "step": 849 }, { "epoch": 0.08, "learning_rate": 0.0019848969521453455, "loss": 1.9355, "step": 850 }, { "epoch": 0.08, "learning_rate": 0.001984841617846313, "loss": 1.8789, "step": 851 }, { "epoch": 0.08, "learning_rate": 0.0019847861831402288, "loss": 1.9326, "step": 852 }, { "epoch": 0.08, "learning_rate": 0.001984730648032745, "loss": 1.9482, "step": 853 }, { "epoch": 0.08, "learning_rate": 0.001984675012529524, "loss": 1.9092, "step": 854 }, { "epoch": 0.08, "learning_rate": 0.0019846192766362365, "loss": 2.0391, "step": 855 }, { "epoch": 0.08, "learning_rate": 0.0019845634403585664, "loss": 1.958, "step": 856 }, { "epoch": 0.08, "learning_rate": 0.0019845075037022058, "loss": 1.9404, "step": 857 }, { "epoch": 0.08, "learning_rate": 0.001984451466672858, "loss": 1.8535, "step": 858 }, { "epoch": 0.08, "learning_rate": 0.001984395329276235, "loss": 2.0332, "step": 859 }, { "epoch": 0.08, "learning_rate": 0.001984339091518061, "loss": 1.957, "step": 860 }, { "epoch": 0.08, "learning_rate": 0.0019842827534040696, "loss": 2.0166, "step": 861 }, { "epoch": 0.08, "learning_rate": 0.0019842263149400046, "loss": 2.0098, "step": 862 }, { "epoch": 0.09, "learning_rate": 0.0019841697761316195, "loss": 1.8682, "step": 863 }, { "epoch": 0.09, "learning_rate": 0.001984113136984679, "loss": 1.9648, "step": 864 }, { "epoch": 0.09, "learning_rate": 0.001984056397504958, "loss": 2.0195, "step": 865 }, { "epoch": 0.09, "learning_rate": 0.00198399955769824, "loss": 1.9941, "step": 866 }, { "epoch": 0.09, "learning_rate": 0.0019839426175703215, "loss": 1.9111, "step": 867 }, { "epoch": 0.09, "learning_rate": 0.0019838855771270067, "loss": 1.9277, "step": 868 }, { "epoch": 0.09, "learning_rate": 0.001983828436374111, "loss": 1.9424, "step": 869 }, { "epoch": 0.09, "learning_rate": 0.0019837711953174606, "loss": 1.9004, "step": 870 }, { "epoch": 0.09, "learning_rate": 0.001983713853962891, "loss": 2.0264, "step": 871 }, { "epoch": 0.09, "learning_rate": 0.0019836564123162483, "loss": 1.8887, "step": 872 }, { "epoch": 0.09, "learning_rate": 0.001983598870383389, "loss": 1.9287, "step": 873 }, { "epoch": 0.09, "learning_rate": 0.0019835412281701795, "loss": 1.9541, "step": 874 }, { "epoch": 0.09, "learning_rate": 0.0019834834856824967, "loss": 2.0352, "step": 875 }, { "epoch": 0.09, "learning_rate": 0.001983425642926227, "loss": 2.0117, "step": 876 }, { "epoch": 0.09, "learning_rate": 0.0019833676999072686, "loss": 2.0, "step": 877 }, { "epoch": 0.09, "learning_rate": 0.001983309656631528, "loss": 2.0146, "step": 878 }, { "epoch": 0.09, "learning_rate": 0.0019832515131049237, "loss": 2.0459, "step": 879 }, { "epoch": 0.09, "learning_rate": 0.0019831932693333827, "loss": 1.9648, "step": 880 }, { "epoch": 0.09, "learning_rate": 0.001983134925322844, "loss": 2.0166, "step": 881 }, { "epoch": 0.09, "learning_rate": 0.0019830764810792553, "loss": 1.9482, "step": 882 }, { "epoch": 0.09, "learning_rate": 0.001983017936608575, "loss": 1.9121, "step": 883 }, { "epoch": 0.09, "learning_rate": 0.0019829592919167723, "loss": 2.002, "step": 884 }, { "epoch": 0.09, "learning_rate": 0.0019829005470098264, "loss": 1.9561, "step": 885 }, { "epoch": 0.09, "learning_rate": 0.001982841701893726, "loss": 2.0332, "step": 886 }, { "epoch": 0.09, "learning_rate": 0.0019827827565744704, "loss": 1.8945, "step": 887 }, { "epoch": 0.09, "learning_rate": 0.0019827237110580697, "loss": 1.9463, "step": 888 }, { "epoch": 0.09, "learning_rate": 0.001982664565350543, "loss": 1.9326, "step": 889 }, { "epoch": 0.09, "learning_rate": 0.0019826053194579216, "loss": 1.9326, "step": 890 }, { "epoch": 0.09, "learning_rate": 0.0019825459733862445, "loss": 1.9326, "step": 891 }, { "epoch": 0.09, "learning_rate": 0.0019824865271415625, "loss": 1.9111, "step": 892 }, { "epoch": 0.09, "learning_rate": 0.0019824269807299367, "loss": 2.0146, "step": 893 }, { "epoch": 0.09, "learning_rate": 0.0019823673341574383, "loss": 1.917, "step": 894 }, { "epoch": 0.09, "learning_rate": 0.0019823075874301475, "loss": 1.9863, "step": 895 }, { "epoch": 0.09, "learning_rate": 0.0019822477405541557, "loss": 1.9639, "step": 896 }, { "epoch": 0.09, "learning_rate": 0.001982187793535565, "loss": 1.9756, "step": 897 }, { "epoch": 0.09, "learning_rate": 0.001982127746380487, "loss": 1.9814, "step": 898 }, { "epoch": 0.09, "learning_rate": 0.0019820675990950432, "loss": 1.9697, "step": 899 }, { "epoch": 0.09, "learning_rate": 0.0019820073516853666, "loss": 2.0596, "step": 900 }, { "epoch": 0.09, "learning_rate": 0.001981947004157599, "loss": 1.9775, "step": 901 }, { "epoch": 0.09, "learning_rate": 0.001981886556517893, "loss": 1.8594, "step": 902 }, { "epoch": 0.09, "learning_rate": 0.0019818260087724115, "loss": 1.8896, "step": 903 }, { "epoch": 0.09, "learning_rate": 0.0019817653609273274, "loss": 2.0166, "step": 904 }, { "epoch": 0.09, "learning_rate": 0.001981704612988824, "loss": 1.9717, "step": 905 }, { "epoch": 0.09, "learning_rate": 0.0019816437649630946, "loss": 1.8545, "step": 906 }, { "epoch": 0.09, "learning_rate": 0.0019815828168563434, "loss": 1.9404, "step": 907 }, { "epoch": 0.09, "learning_rate": 0.001981521768674783, "loss": 1.9619, "step": 908 }, { "epoch": 0.09, "learning_rate": 0.0019814606204246387, "loss": 2.0254, "step": 909 }, { "epoch": 0.09, "learning_rate": 0.0019813993721121436, "loss": 1.9932, "step": 910 }, { "epoch": 0.09, "learning_rate": 0.0019813380237435427, "loss": 2.0596, "step": 911 }, { "epoch": 0.09, "learning_rate": 0.0019812765753250907, "loss": 1.9854, "step": 912 }, { "epoch": 0.09, "learning_rate": 0.0019812150268630523, "loss": 1.9668, "step": 913 }, { "epoch": 0.09, "learning_rate": 0.0019811533783637026, "loss": 1.9971, "step": 914 }, { "epoch": 0.09, "learning_rate": 0.0019810916298333265, "loss": 2.0215, "step": 915 }, { "epoch": 0.09, "learning_rate": 0.0019810297812782192, "loss": 1.9453, "step": 916 }, { "epoch": 0.09, "learning_rate": 0.0019809678327046875, "loss": 1.999, "step": 917 }, { "epoch": 0.09, "learning_rate": 0.001980905784119046, "loss": 2.0635, "step": 918 }, { "epoch": 0.09, "learning_rate": 0.0019808436355276218, "loss": 1.9707, "step": 919 }, { "epoch": 0.09, "learning_rate": 0.0019807813869367496, "loss": 1.8145, "step": 920 }, { "epoch": 0.09, "learning_rate": 0.0019807190383527773, "loss": 1.9414, "step": 921 }, { "epoch": 0.09, "learning_rate": 0.0019806565897820603, "loss": 1.9922, "step": 922 }, { "epoch": 0.09, "learning_rate": 0.0019805940412309666, "loss": 1.9834, "step": 923 }, { "epoch": 0.09, "learning_rate": 0.001980531392705872, "loss": 2.0059, "step": 924 }, { "epoch": 0.09, "learning_rate": 0.001980468644213164, "loss": 1.9043, "step": 925 }, { "epoch": 0.09, "learning_rate": 0.001980405795759241, "loss": 1.8857, "step": 926 }, { "epoch": 0.09, "learning_rate": 0.001980342847350509, "loss": 1.9854, "step": 927 }, { "epoch": 0.09, "learning_rate": 0.001980279798993387, "loss": 1.9902, "step": 928 }, { "epoch": 0.09, "learning_rate": 0.0019802166506943017, "loss": 1.9668, "step": 929 }, { "epoch": 0.09, "learning_rate": 0.0019801534024596924, "loss": 2.0, "step": 930 }, { "epoch": 0.09, "learning_rate": 0.001980090054296007, "loss": 1.915, "step": 931 }, { "epoch": 0.09, "learning_rate": 0.0019800266062097037, "loss": 1.9395, "step": 932 }, { "epoch": 0.09, "learning_rate": 0.0019799630582072515, "loss": 1.9814, "step": 933 }, { "epoch": 0.09, "learning_rate": 0.0019798994102951293, "loss": 2.0205, "step": 934 }, { "epoch": 0.09, "learning_rate": 0.001979835662479826, "loss": 1.9238, "step": 935 }, { "epoch": 0.09, "learning_rate": 0.001979771814767841, "loss": 1.877, "step": 936 }, { "epoch": 0.09, "learning_rate": 0.0019797078671656833, "loss": 1.998, "step": 937 }, { "epoch": 0.09, "learning_rate": 0.001979643819679873, "loss": 1.9189, "step": 938 }, { "epoch": 0.09, "learning_rate": 0.00197957967231694, "loss": 2.0107, "step": 939 }, { "epoch": 0.09, "learning_rate": 0.001979515425083424, "loss": 1.9717, "step": 940 }, { "epoch": 0.09, "learning_rate": 0.0019794510779858753, "loss": 1.9873, "step": 941 }, { "epoch": 0.09, "learning_rate": 0.001979386631030854, "loss": 1.9111, "step": 942 }, { "epoch": 0.09, "learning_rate": 0.001979322084224931, "loss": 1.8965, "step": 943 }, { "epoch": 0.09, "learning_rate": 0.0019792574375746864, "loss": 1.957, "step": 944 }, { "epoch": 0.09, "learning_rate": 0.0019791926910867116, "loss": 1.9375, "step": 945 }, { "epoch": 0.09, "learning_rate": 0.001979127844767608, "loss": 2.002, "step": 946 }, { "epoch": 0.09, "learning_rate": 0.0019790628986239857, "loss": 1.9863, "step": 947 }, { "epoch": 0.09, "learning_rate": 0.0019789978526624672, "loss": 2.0498, "step": 948 }, { "epoch": 0.09, "learning_rate": 0.001978932706889684, "loss": 1.9443, "step": 949 }, { "epoch": 0.09, "learning_rate": 0.001978867461312277, "loss": 1.9531, "step": 950 }, { "epoch": 0.09, "learning_rate": 0.0019788021159368994, "loss": 1.9561, "step": 951 }, { "epoch": 0.09, "learning_rate": 0.001978736670770212, "loss": 1.8877, "step": 952 }, { "epoch": 0.09, "learning_rate": 0.0019786711258188883, "loss": 2.0186, "step": 953 }, { "epoch": 0.09, "learning_rate": 0.00197860548108961, "loss": 1.9463, "step": 954 }, { "epoch": 0.09, "learning_rate": 0.0019785397365890698, "loss": 1.9766, "step": 955 }, { "epoch": 0.09, "learning_rate": 0.001978473892323971, "loss": 1.9951, "step": 956 }, { "epoch": 0.09, "learning_rate": 0.001978407948301026, "loss": 2.0146, "step": 957 }, { "epoch": 0.09, "learning_rate": 0.0019783419045269583, "loss": 1.9111, "step": 958 }, { "epoch": 0.09, "learning_rate": 0.0019782757610085013, "loss": 1.9326, "step": 959 }, { "epoch": 0.09, "learning_rate": 0.001978209517752398, "loss": 2.0127, "step": 960 }, { "epoch": 0.09, "learning_rate": 0.0019781431747654026, "loss": 1.9717, "step": 961 }, { "epoch": 0.09, "learning_rate": 0.001978076732054279, "loss": 1.9209, "step": 962 }, { "epoch": 0.09, "learning_rate": 0.001978010189625801, "loss": 1.8516, "step": 963 }, { "epoch": 0.1, "learning_rate": 0.0019779435474867524, "loss": 2.0117, "step": 964 }, { "epoch": 0.1, "learning_rate": 0.0019778768056439274, "loss": 1.9912, "step": 965 }, { "epoch": 0.1, "learning_rate": 0.0019778099641041317, "loss": 1.9365, "step": 966 }, { "epoch": 0.1, "learning_rate": 0.001977743022874179, "loss": 1.9746, "step": 967 }, { "epoch": 0.1, "learning_rate": 0.0019776759819608945, "loss": 1.957, "step": 968 }, { "epoch": 0.1, "learning_rate": 0.0019776088413711126, "loss": 1.9199, "step": 969 }, { "epoch": 0.1, "learning_rate": 0.0019775416011116793, "loss": 2.0176, "step": 970 }, { "epoch": 0.1, "learning_rate": 0.001977474261189449, "loss": 1.8955, "step": 971 }, { "epoch": 0.1, "learning_rate": 0.001977406821611288, "loss": 1.9707, "step": 972 }, { "epoch": 0.1, "learning_rate": 0.0019773392823840713, "loss": 1.9492, "step": 973 }, { "epoch": 0.1, "learning_rate": 0.001977271643514685, "loss": 2.0098, "step": 974 }, { "epoch": 0.1, "learning_rate": 0.0019772039050100255, "loss": 1.9775, "step": 975 }, { "epoch": 0.1, "learning_rate": 0.001977136066876998, "loss": 1.957, "step": 976 }, { "epoch": 0.1, "learning_rate": 0.0019770681291225193, "loss": 1.9736, "step": 977 }, { "epoch": 0.1, "learning_rate": 0.001977000091753516, "loss": 1.9521, "step": 978 }, { "epoch": 0.1, "learning_rate": 0.001976931954776924, "loss": 1.8965, "step": 979 }, { "epoch": 0.1, "learning_rate": 0.001976863718199691, "loss": 1.9473, "step": 980 }, { "epoch": 0.1, "learning_rate": 0.0019767953820287727, "loss": 1.958, "step": 981 }, { "epoch": 0.1, "learning_rate": 0.001976726946271137, "loss": 1.9541, "step": 982 }, { "epoch": 0.1, "learning_rate": 0.001976658410933761, "loss": 2.0186, "step": 983 }, { "epoch": 0.1, "learning_rate": 0.0019765897760236316, "loss": 1.9072, "step": 984 }, { "epoch": 0.1, "learning_rate": 0.0019765210415477468, "loss": 1.8955, "step": 985 }, { "epoch": 0.1, "learning_rate": 0.001976452207513114, "loss": 1.9199, "step": 986 }, { "epoch": 0.1, "learning_rate": 0.001976383273926751, "loss": 1.9824, "step": 987 }, { "epoch": 0.1, "learning_rate": 0.001976314240795686, "loss": 1.9111, "step": 988 }, { "epoch": 0.1, "learning_rate": 0.0019762451081269568, "loss": 1.9668, "step": 989 }, { "epoch": 0.1, "learning_rate": 0.001976175875927612, "loss": 1.8779, "step": 990 }, { "epoch": 0.1, "learning_rate": 0.0019761065442047096, "loss": 1.8418, "step": 991 }, { "epoch": 0.1, "learning_rate": 0.0019760371129653185, "loss": 1.9072, "step": 992 }, { "epoch": 0.1, "learning_rate": 0.0019759675822165167, "loss": 1.9424, "step": 993 }, { "epoch": 0.1, "learning_rate": 0.001975897951965394, "loss": 1.9971, "step": 994 }, { "epoch": 0.1, "learning_rate": 0.0019758282222190485, "loss": 1.9307, "step": 995 }, { "epoch": 0.1, "learning_rate": 0.0019757583929845895, "loss": 1.9004, "step": 996 }, { "epoch": 0.1, "learning_rate": 0.001975688464269137, "loss": 1.8867, "step": 997 }, { "epoch": 0.1, "learning_rate": 0.00197561843607982, "loss": 1.9795, "step": 998 }, { "epoch": 0.1, "learning_rate": 0.0019755483084237776, "loss": 1.916, "step": 999 }, { "epoch": 0.1, "learning_rate": 0.00197547808130816, "loss": 1.9824, "step": 1000 }, { "epoch": 0.1, "learning_rate": 0.001975407754740127, "loss": 1.9639, "step": 1001 }, { "epoch": 0.1, "learning_rate": 0.001975337328726848, "loss": 2.0068, "step": 1002 }, { "epoch": 0.1, "learning_rate": 0.0019752668032755037, "loss": 2.0068, "step": 1003 }, { "epoch": 0.1, "learning_rate": 0.0019751961783932843, "loss": 1.8037, "step": 1004 }, { "epoch": 0.1, "learning_rate": 0.0019751254540873895, "loss": 1.8672, "step": 1005 }, { "epoch": 0.1, "learning_rate": 0.0019750546303650314, "loss": 1.9229, "step": 1006 }, { "epoch": 0.1, "learning_rate": 0.0019749837072334285, "loss": 1.9961, "step": 1007 }, { "epoch": 0.1, "learning_rate": 0.0019749126846998134, "loss": 1.958, "step": 1008 }, { "epoch": 0.1, "learning_rate": 0.0019748415627714264, "loss": 1.9951, "step": 1009 }, { "epoch": 0.1, "learning_rate": 0.0019747703414555186, "loss": 1.9883, "step": 1010 }, { "epoch": 0.1, "learning_rate": 0.0019746990207593503, "loss": 2.0137, "step": 1011 }, { "epoch": 0.1, "learning_rate": 0.0019746276006901943, "loss": 1.8613, "step": 1012 }, { "epoch": 0.1, "learning_rate": 0.001974556081255331, "loss": 1.9854, "step": 1013 }, { "epoch": 0.1, "learning_rate": 0.001974484462462053, "loss": 1.8818, "step": 1014 }, { "epoch": 0.1, "learning_rate": 0.001974412744317661, "loss": 1.9336, "step": 1015 }, { "epoch": 0.1, "learning_rate": 0.0019743409268294667, "loss": 1.9678, "step": 1016 }, { "epoch": 0.1, "learning_rate": 0.0019742690100047933, "loss": 1.875, "step": 1017 }, { "epoch": 0.1, "learning_rate": 0.0019741969938509717, "loss": 1.9375, "step": 1018 }, { "epoch": 0.1, "learning_rate": 0.001974124878375345, "loss": 1.8516, "step": 1019 }, { "epoch": 0.1, "learning_rate": 0.001974052663585265, "loss": 1.9365, "step": 1020 }, { "epoch": 0.1, "learning_rate": 0.001973980349488094, "loss": 1.9707, "step": 1021 }, { "epoch": 0.1, "learning_rate": 0.0019739079360912054, "loss": 1.9863, "step": 1022 }, { "epoch": 0.1, "learning_rate": 0.001973835423401981, "loss": 1.959, "step": 1023 }, { "epoch": 0.1, "learning_rate": 0.0019737628114278147, "loss": 1.96, "step": 1024 }, { "epoch": 0.1, "learning_rate": 0.0019736901001761085, "loss": 1.9795, "step": 1025 }, { "epoch": 0.1, "learning_rate": 0.001973617289654276, "loss": 1.8779, "step": 1026 }, { "epoch": 0.1, "learning_rate": 0.00197354437986974, "loss": 1.9561, "step": 1027 }, { "epoch": 0.1, "learning_rate": 0.0019734713708299345, "loss": 2.041, "step": 1028 }, { "epoch": 0.1, "learning_rate": 0.0019733982625423025, "loss": 1.9102, "step": 1029 }, { "epoch": 0.1, "learning_rate": 0.0019733250550142975, "loss": 1.9404, "step": 1030 }, { "epoch": 0.1, "learning_rate": 0.0019732517482533834, "loss": 1.9102, "step": 1031 }, { "epoch": 0.1, "learning_rate": 0.001973178342267034, "loss": 1.9678, "step": 1032 }, { "epoch": 0.1, "learning_rate": 0.001973104837062733, "loss": 1.9346, "step": 1033 }, { "epoch": 0.1, "learning_rate": 0.0019730312326479748, "loss": 2.043, "step": 1034 }, { "epoch": 0.1, "learning_rate": 0.001972957529030263, "loss": 1.9883, "step": 1035 }, { "epoch": 0.1, "learning_rate": 0.001972883726217113, "loss": 1.9111, "step": 1036 }, { "epoch": 0.1, "learning_rate": 0.0019728098242160477, "loss": 1.8398, "step": 1037 }, { "epoch": 0.1, "learning_rate": 0.0019727358230346025, "loss": 1.9346, "step": 1038 }, { "epoch": 0.1, "learning_rate": 0.001972661722680322, "loss": 1.9238, "step": 1039 }, { "epoch": 0.1, "learning_rate": 0.0019725875231607604, "loss": 1.9688, "step": 1040 }, { "epoch": 0.1, "learning_rate": 0.001972513224483483, "loss": 1.9609, "step": 1041 }, { "epoch": 0.1, "learning_rate": 0.0019724388266560643, "loss": 1.8838, "step": 1042 }, { "epoch": 0.1, "learning_rate": 0.0019723643296860904, "loss": 1.9961, "step": 1043 }, { "epoch": 0.1, "learning_rate": 0.001972289733581155, "loss": 1.9736, "step": 1044 }, { "epoch": 0.1, "learning_rate": 0.001972215038348864, "loss": 1.9121, "step": 1045 }, { "epoch": 0.1, "learning_rate": 0.001972140243996833, "loss": 1.8799, "step": 1046 }, { "epoch": 0.1, "learning_rate": 0.0019720653505326874, "loss": 1.8984, "step": 1047 }, { "epoch": 0.1, "learning_rate": 0.0019719903579640623, "loss": 1.9316, "step": 1048 }, { "epoch": 0.1, "learning_rate": 0.001971915266298604, "loss": 1.9404, "step": 1049 }, { "epoch": 0.1, "learning_rate": 0.0019718400755439677, "loss": 1.9971, "step": 1050 }, { "epoch": 0.1, "learning_rate": 0.00197176478570782, "loss": 1.8984, "step": 1051 }, { "epoch": 0.1, "learning_rate": 0.0019716893967978363, "loss": 1.9131, "step": 1052 }, { "epoch": 0.1, "learning_rate": 0.001971613908821703, "loss": 1.9209, "step": 1053 }, { "epoch": 0.1, "learning_rate": 0.0019715383217871157, "loss": 1.9697, "step": 1054 }, { "epoch": 0.1, "learning_rate": 0.0019714626357017814, "loss": 2.0088, "step": 1055 }, { "epoch": 0.1, "learning_rate": 0.0019713868505734163, "loss": 1.8945, "step": 1056 }, { "epoch": 0.1, "learning_rate": 0.0019713109664097465, "loss": 1.8682, "step": 1057 }, { "epoch": 0.1, "learning_rate": 0.0019712349832185093, "loss": 1.9395, "step": 1058 }, { "epoch": 0.1, "learning_rate": 0.0019711589010074507, "loss": 1.873, "step": 1059 }, { "epoch": 0.1, "learning_rate": 0.001971082719784328, "loss": 1.9854, "step": 1060 }, { "epoch": 0.1, "learning_rate": 0.0019710064395569073, "loss": 1.9268, "step": 1061 }, { "epoch": 0.1, "learning_rate": 0.001970930060332966, "loss": 1.9258, "step": 1062 }, { "epoch": 0.1, "learning_rate": 0.001970853582120292, "loss": 1.8398, "step": 1063 }, { "epoch": 0.1, "learning_rate": 0.001970777004926681, "loss": 1.9775, "step": 1064 }, { "epoch": 0.1, "learning_rate": 0.001970700328759941, "loss": 1.9648, "step": 1065 }, { "epoch": 0.11, "learning_rate": 0.001970623553627889, "loss": 1.9697, "step": 1066 }, { "epoch": 0.11, "learning_rate": 0.0019705466795383528, "loss": 1.832, "step": 1067 }, { "epoch": 0.11, "learning_rate": 0.00197046970649917, "loss": 1.9629, "step": 1068 }, { "epoch": 0.11, "learning_rate": 0.0019703926345181876, "loss": 1.9053, "step": 1069 }, { "epoch": 0.11, "learning_rate": 0.0019703154636032637, "loss": 1.8643, "step": 1070 }, { "epoch": 0.11, "learning_rate": 0.001970238193762266, "loss": 1.9229, "step": 1071 }, { "epoch": 0.11, "learning_rate": 0.0019701608250030722, "loss": 1.8828, "step": 1072 }, { "epoch": 0.11, "learning_rate": 0.001970083357333571, "loss": 1.9688, "step": 1073 }, { "epoch": 0.11, "learning_rate": 0.0019700057907616588, "loss": 1.9531, "step": 1074 }, { "epoch": 0.11, "learning_rate": 0.001969928125295245, "loss": 1.7979, "step": 1075 }, { "epoch": 0.11, "learning_rate": 0.0019698503609422474, "loss": 1.9189, "step": 1076 }, { "epoch": 0.11, "learning_rate": 0.0019697724977105944, "loss": 1.9795, "step": 1077 }, { "epoch": 0.11, "learning_rate": 0.0019696945356082246, "loss": 1.9971, "step": 1078 }, { "epoch": 0.11, "learning_rate": 0.0019696164746430857, "loss": 1.9258, "step": 1079 }, { "epoch": 0.11, "learning_rate": 0.001969538314823137, "loss": 2.0293, "step": 1080 }, { "epoch": 0.11, "learning_rate": 0.0019694600561563466, "loss": 1.8389, "step": 1081 }, { "epoch": 0.11, "learning_rate": 0.001969381698650693, "loss": 1.9482, "step": 1082 }, { "epoch": 0.11, "learning_rate": 0.001969303242314165, "loss": 2.0479, "step": 1083 }, { "epoch": 0.11, "learning_rate": 0.0019692246871547627, "loss": 1.835, "step": 1084 }, { "epoch": 0.11, "learning_rate": 0.001969146033180493, "loss": 1.9131, "step": 1085 }, { "epoch": 0.11, "learning_rate": 0.001969067280399376, "loss": 1.8887, "step": 1086 }, { "epoch": 0.11, "learning_rate": 0.001968988428819441, "loss": 1.8574, "step": 1087 }, { "epoch": 0.11, "learning_rate": 0.001968909478448726, "loss": 1.9248, "step": 1088 }, { "epoch": 0.11, "learning_rate": 0.001968830429295281, "loss": 2.0225, "step": 1089 }, { "epoch": 0.11, "learning_rate": 0.001968751281367165, "loss": 1.9609, "step": 1090 }, { "epoch": 0.11, "learning_rate": 0.001968672034672448, "loss": 1.9316, "step": 1091 }, { "epoch": 0.11, "learning_rate": 0.0019685926892192085, "loss": 1.8359, "step": 1092 }, { "epoch": 0.11, "learning_rate": 0.0019685132450155364, "loss": 2.0166, "step": 1093 }, { "epoch": 0.11, "learning_rate": 0.001968433702069531, "loss": 1.9834, "step": 1094 }, { "epoch": 0.11, "learning_rate": 0.001968354060389302, "loss": 1.9033, "step": 1095 }, { "epoch": 0.11, "learning_rate": 0.0019682743199829693, "loss": 1.9131, "step": 1096 }, { "epoch": 0.11, "learning_rate": 0.0019681944808586623, "loss": 1.8916, "step": 1097 }, { "epoch": 0.11, "learning_rate": 0.0019681145430245207, "loss": 1.9277, "step": 1098 }, { "epoch": 0.11, "learning_rate": 0.0019680345064886948, "loss": 1.9688, "step": 1099 }, { "epoch": 0.11, "learning_rate": 0.0019679543712593445, "loss": 1.9453, "step": 1100 }, { "epoch": 0.11, "learning_rate": 0.0019678741373446392, "loss": 1.9316, "step": 1101 }, { "epoch": 0.11, "learning_rate": 0.0019677938047527595, "loss": 1.9551, "step": 1102 }, { "epoch": 0.11, "learning_rate": 0.0019677133734918957, "loss": 1.9072, "step": 1103 }, { "epoch": 0.11, "learning_rate": 0.001967632843570247, "loss": 1.8643, "step": 1104 }, { "epoch": 0.11, "learning_rate": 0.001967552214996025, "loss": 2.0615, "step": 1105 }, { "epoch": 0.11, "learning_rate": 0.0019674714877774484, "loss": 1.9316, "step": 1106 }, { "epoch": 0.11, "learning_rate": 0.0019673906619227487, "loss": 1.8965, "step": 1107 }, { "epoch": 0.11, "learning_rate": 0.001967309737440166, "loss": 1.9746, "step": 1108 }, { "epoch": 0.11, "learning_rate": 0.001967228714337951, "loss": 1.8926, "step": 1109 }, { "epoch": 0.11, "learning_rate": 0.0019671475926243635, "loss": 1.8896, "step": 1110 }, { "epoch": 0.11, "learning_rate": 0.001967066372307675, "loss": 1.7705, "step": 1111 }, { "epoch": 0.11, "learning_rate": 0.001966985053396165, "loss": 1.915, "step": 1112 }, { "epoch": 0.11, "learning_rate": 0.0019669036358981253, "loss": 1.9092, "step": 1113 }, { "epoch": 0.11, "learning_rate": 0.0019668221198218556, "loss": 1.9092, "step": 1114 }, { "epoch": 0.11, "learning_rate": 0.001966740505175667, "loss": 1.9873, "step": 1115 }, { "epoch": 0.11, "learning_rate": 0.0019666587919678815, "loss": 1.8721, "step": 1116 }, { "epoch": 0.11, "learning_rate": 0.001966576980206828, "loss": 1.9668, "step": 1117 }, { "epoch": 0.11, "learning_rate": 0.0019664950699008485, "loss": 1.9541, "step": 1118 }, { "epoch": 0.11, "learning_rate": 0.0019664130610582936, "loss": 1.7725, "step": 1119 }, { "epoch": 0.11, "learning_rate": 0.0019663309536875253, "loss": 1.8447, "step": 1120 }, { "epoch": 0.11, "learning_rate": 0.0019662487477969132, "loss": 1.9072, "step": 1121 }, { "epoch": 0.11, "learning_rate": 0.001966166443394839, "loss": 1.8623, "step": 1122 }, { "epoch": 0.11, "learning_rate": 0.0019660840404896944, "loss": 1.8164, "step": 1123 }, { "epoch": 0.11, "learning_rate": 0.0019660015390898797, "loss": 1.874, "step": 1124 }, { "epoch": 0.11, "learning_rate": 0.0019659189392038064, "loss": 1.9355, "step": 1125 }, { "epoch": 0.11, "learning_rate": 0.001965836240839896, "loss": 2.0166, "step": 1126 }, { "epoch": 0.11, "learning_rate": 0.0019657534440065795, "loss": 1.8926, "step": 1127 }, { "epoch": 0.11, "learning_rate": 0.0019656705487122988, "loss": 1.8896, "step": 1128 }, { "epoch": 0.11, "learning_rate": 0.0019655875549655045, "loss": 1.9033, "step": 1129 }, { "epoch": 0.11, "learning_rate": 0.0019655044627746587, "loss": 1.9785, "step": 1130 }, { "epoch": 0.11, "learning_rate": 0.001965421272148232, "loss": 1.9102, "step": 1131 }, { "epoch": 0.11, "learning_rate": 0.0019653379830947074, "loss": 1.9072, "step": 1132 }, { "epoch": 0.11, "learning_rate": 0.0019652545956225747, "loss": 1.8613, "step": 1133 }, { "epoch": 0.11, "learning_rate": 0.0019651711097403365, "loss": 1.8398, "step": 1134 }, { "epoch": 0.11, "learning_rate": 0.001965087525456504, "loss": 1.9678, "step": 1135 }, { "epoch": 0.11, "learning_rate": 0.0019650038427795995, "loss": 1.9238, "step": 1136 }, { "epoch": 0.11, "learning_rate": 0.0019649200617181534, "loss": 1.9775, "step": 1137 }, { "epoch": 0.11, "learning_rate": 0.0019648361822807085, "loss": 1.9414, "step": 1138 }, { "epoch": 0.11, "learning_rate": 0.001964752204475816, "loss": 1.793, "step": 1139 }, { "epoch": 0.11, "learning_rate": 0.0019646681283120374, "loss": 1.9336, "step": 1140 }, { "epoch": 0.11, "learning_rate": 0.0019645839537979454, "loss": 1.9072, "step": 1141 }, { "epoch": 0.11, "learning_rate": 0.001964499680942121, "loss": 1.8975, "step": 1142 }, { "epoch": 0.11, "learning_rate": 0.0019644153097531565, "loss": 1.7832, "step": 1143 }, { "epoch": 0.11, "learning_rate": 0.001964330840239653, "loss": 1.8867, "step": 1144 }, { "epoch": 0.11, "learning_rate": 0.001964246272410223, "loss": 1.8496, "step": 1145 }, { "epoch": 0.11, "learning_rate": 0.0019641616062734887, "loss": 1.9658, "step": 1146 }, { "epoch": 0.11, "learning_rate": 0.0019640768418380814, "loss": 1.8916, "step": 1147 }, { "epoch": 0.11, "learning_rate": 0.0019639919791126434, "loss": 1.9248, "step": 1148 }, { "epoch": 0.11, "learning_rate": 0.0019639070181058263, "loss": 1.8711, "step": 1149 }, { "epoch": 0.11, "learning_rate": 0.0019638219588262924, "loss": 1.8164, "step": 1150 }, { "epoch": 0.11, "learning_rate": 0.0019637368012827136, "loss": 1.8994, "step": 1151 }, { "epoch": 0.11, "learning_rate": 0.001963651545483772, "loss": 2.0137, "step": 1152 }, { "epoch": 0.11, "learning_rate": 0.0019635661914381595, "loss": 1.8975, "step": 1153 }, { "epoch": 0.11, "learning_rate": 0.0019634807391545783, "loss": 1.832, "step": 1154 }, { "epoch": 0.11, "learning_rate": 0.0019633951886417404, "loss": 1.9697, "step": 1155 }, { "epoch": 0.11, "learning_rate": 0.001963309539908368, "loss": 1.8369, "step": 1156 }, { "epoch": 0.11, "learning_rate": 0.001963223792963193, "loss": 1.8906, "step": 1157 }, { "epoch": 0.11, "learning_rate": 0.0019631379478149574, "loss": 1.9033, "step": 1158 }, { "epoch": 0.11, "learning_rate": 0.001963052004472414, "loss": 1.9834, "step": 1159 }, { "epoch": 0.11, "learning_rate": 0.001962965962944324, "loss": 1.9229, "step": 1160 }, { "epoch": 0.11, "learning_rate": 0.0019628798232394604, "loss": 1.8428, "step": 1161 }, { "epoch": 0.11, "learning_rate": 0.001962793585366605, "loss": 1.9756, "step": 1162 }, { "epoch": 0.11, "learning_rate": 0.0019627072493345497, "loss": 1.9404, "step": 1163 }, { "epoch": 0.11, "learning_rate": 0.0019626208151520973, "loss": 1.9697, "step": 1164 }, { "epoch": 0.11, "learning_rate": 0.001962534282828059, "loss": 1.8955, "step": 1165 }, { "epoch": 0.11, "learning_rate": 0.001962447652371258, "loss": 1.873, "step": 1166 }, { "epoch": 0.12, "learning_rate": 0.001962360923790526, "loss": 2.0059, "step": 1167 }, { "epoch": 0.12, "learning_rate": 0.001962274097094705, "loss": 2.002, "step": 1168 }, { "epoch": 0.12, "learning_rate": 0.0019621871722926482, "loss": 1.9346, "step": 1169 }, { "epoch": 0.12, "learning_rate": 0.0019621001493932166, "loss": 1.9834, "step": 1170 }, { "epoch": 0.12, "learning_rate": 0.0019620130284052828, "loss": 1.9893, "step": 1171 }, { "epoch": 0.12, "learning_rate": 0.0019619258093377294, "loss": 1.9609, "step": 1172 }, { "epoch": 0.12, "learning_rate": 0.0019618384921994475, "loss": 1.8926, "step": 1173 }, { "epoch": 0.12, "learning_rate": 0.001961751076999341, "loss": 1.9326, "step": 1174 }, { "epoch": 0.12, "learning_rate": 0.0019616635637463207, "loss": 1.8506, "step": 1175 }, { "epoch": 0.12, "learning_rate": 0.001961575952449309, "loss": 1.8848, "step": 1176 }, { "epoch": 0.12, "learning_rate": 0.001961488243117239, "loss": 1.9209, "step": 1177 }, { "epoch": 0.12, "learning_rate": 0.001961400435759052, "loss": 1.9541, "step": 1178 }, { "epoch": 0.12, "learning_rate": 0.0019613125303837003, "loss": 1.8564, "step": 1179 }, { "epoch": 0.12, "learning_rate": 0.0019612245270001463, "loss": 1.9404, "step": 1180 }, { "epoch": 0.12, "learning_rate": 0.001961136425617362, "loss": 1.9102, "step": 1181 }, { "epoch": 0.12, "learning_rate": 0.0019610482262443297, "loss": 1.9062, "step": 1182 }, { "epoch": 0.12, "learning_rate": 0.0019609599288900412, "loss": 2.0137, "step": 1183 }, { "epoch": 0.12, "learning_rate": 0.001960871533563499, "loss": 1.9961, "step": 1184 }, { "epoch": 0.12, "learning_rate": 0.001960783040273715, "loss": 1.96, "step": 1185 }, { "epoch": 0.12, "learning_rate": 0.0019606944490297117, "loss": 1.9199, "step": 1186 }, { "epoch": 0.12, "learning_rate": 0.001960605759840521, "loss": 1.9258, "step": 1187 }, { "epoch": 0.12, "learning_rate": 0.0019605169727151847, "loss": 1.9873, "step": 1188 }, { "epoch": 0.12, "learning_rate": 0.001960428087662755, "loss": 1.8799, "step": 1189 }, { "epoch": 0.12, "learning_rate": 0.0019603391046922936, "loss": 1.8477, "step": 1190 }, { "epoch": 0.12, "learning_rate": 0.0019602500238128737, "loss": 1.834, "step": 1191 }, { "epoch": 0.12, "learning_rate": 0.001960160845033576, "loss": 1.7979, "step": 1192 }, { "epoch": 0.12, "learning_rate": 0.0019600715683634937, "loss": 1.8545, "step": 1193 }, { "epoch": 0.12, "learning_rate": 0.001959982193811728, "loss": 1.8984, "step": 1194 }, { "epoch": 0.12, "learning_rate": 0.0019598927213873908, "loss": 1.8779, "step": 1195 }, { "epoch": 0.12, "learning_rate": 0.001959803151099604, "loss": 1.9678, "step": 1196 }, { "epoch": 0.12, "learning_rate": 0.0019597134829575, "loss": 1.9062, "step": 1197 }, { "epoch": 0.12, "learning_rate": 0.001959623716970221, "loss": 1.8994, "step": 1198 }, { "epoch": 0.12, "learning_rate": 0.0019595338531469175, "loss": 1.9209, "step": 1199 }, { "epoch": 0.12, "learning_rate": 0.0019594438914967522, "loss": 1.9697, "step": 1200 }, { "epoch": 0.12, "learning_rate": 0.0019593538320288977, "loss": 1.8516, "step": 1201 }, { "epoch": 0.12, "learning_rate": 0.001959263674752534, "loss": 1.9385, "step": 1202 }, { "epoch": 0.12, "learning_rate": 0.0019591734196768543, "loss": 1.8799, "step": 1203 }, { "epoch": 0.12, "learning_rate": 0.0019590830668110596, "loss": 1.9395, "step": 1204 }, { "epoch": 0.12, "learning_rate": 0.0019589926161643623, "loss": 1.8711, "step": 1205 }, { "epoch": 0.12, "learning_rate": 0.001958902067745983, "loss": 1.8428, "step": 1206 }, { "epoch": 0.12, "learning_rate": 0.0019588114215651543, "loss": 1.8623, "step": 1207 }, { "epoch": 0.12, "learning_rate": 0.001958720677631117, "loss": 1.8916, "step": 1208 }, { "epoch": 0.12, "learning_rate": 0.001958629835953124, "loss": 1.8408, "step": 1209 }, { "epoch": 0.12, "learning_rate": 0.0019585388965404353, "loss": 1.8467, "step": 1210 }, { "epoch": 0.12, "learning_rate": 0.001958447859402323, "loss": 1.8203, "step": 1211 }, { "epoch": 0.12, "learning_rate": 0.0019583567245480688, "loss": 1.8857, "step": 1212 }, { "epoch": 0.12, "learning_rate": 0.001958265491986964, "loss": 1.957, "step": 1213 }, { "epoch": 0.12, "learning_rate": 0.00195817416172831, "loss": 1.8281, "step": 1214 }, { "epoch": 0.12, "learning_rate": 0.001958082733781418, "loss": 1.8857, "step": 1215 }, { "epoch": 0.12, "learning_rate": 0.0019579912081556096, "loss": 1.9434, "step": 1216 }, { "epoch": 0.12, "learning_rate": 0.0019578995848602153, "loss": 1.9658, "step": 1217 }, { "epoch": 0.12, "learning_rate": 0.001957807863904577, "loss": 1.9463, "step": 1218 }, { "epoch": 0.12, "learning_rate": 0.001957716045298046, "loss": 1.8877, "step": 1219 }, { "epoch": 0.12, "learning_rate": 0.001957624129049983, "loss": 1.8447, "step": 1220 }, { "epoch": 0.12, "learning_rate": 0.0019575321151697597, "loss": 1.9609, "step": 1221 }, { "epoch": 0.12, "learning_rate": 0.0019574400036667565, "loss": 1.7881, "step": 1222 }, { "epoch": 0.12, "learning_rate": 0.001957347794550364, "loss": 2.0068, "step": 1223 }, { "epoch": 0.12, "learning_rate": 0.0019572554878299845, "loss": 1.9385, "step": 1224 }, { "epoch": 0.12, "learning_rate": 0.001957163083515028, "loss": 1.8633, "step": 1225 }, { "epoch": 0.12, "learning_rate": 0.001957070581614915, "loss": 1.9307, "step": 1226 }, { "epoch": 0.12, "learning_rate": 0.0019569779821390775, "loss": 1.8174, "step": 1227 }, { "epoch": 0.12, "learning_rate": 0.0019568852850969554, "loss": 1.8105, "step": 1228 }, { "epoch": 0.12, "learning_rate": 0.0019567924904979996, "loss": 1.7959, "step": 1229 }, { "epoch": 0.12, "learning_rate": 0.0019566995983516704, "loss": 1.9316, "step": 1230 }, { "epoch": 0.12, "learning_rate": 0.001956606608667439, "loss": 1.9561, "step": 1231 }, { "epoch": 0.12, "learning_rate": 0.001956513521454785, "loss": 1.874, "step": 1232 }, { "epoch": 0.12, "learning_rate": 0.0019564203367232, "loss": 1.9502, "step": 1233 }, { "epoch": 0.12, "learning_rate": 0.0019563270544821843, "loss": 1.8926, "step": 1234 }, { "epoch": 0.12, "learning_rate": 0.001956233674741247, "loss": 1.9561, "step": 1235 }, { "epoch": 0.12, "learning_rate": 0.0019561401975099098, "loss": 1.999, "step": 1236 }, { "epoch": 0.12, "learning_rate": 0.001956046622797702, "loss": 2.0, "step": 1237 }, { "epoch": 0.12, "learning_rate": 0.001955952950614164, "loss": 1.79, "step": 1238 }, { "epoch": 0.12, "learning_rate": 0.001955859180968847, "loss": 1.9814, "step": 1239 }, { "epoch": 0.12, "learning_rate": 0.0019557653138713095, "loss": 1.8193, "step": 1240 }, { "epoch": 0.12, "learning_rate": 0.001955671349331122, "loss": 1.8584, "step": 1241 }, { "epoch": 0.12, "learning_rate": 0.0019555772873578645, "loss": 1.8232, "step": 1242 }, { "epoch": 0.12, "learning_rate": 0.001955483127961127, "loss": 1.8809, "step": 1243 }, { "epoch": 0.12, "learning_rate": 0.001955388871150509, "loss": 2.0088, "step": 1244 }, { "epoch": 0.12, "learning_rate": 0.001955294516935621, "loss": 1.9492, "step": 1245 }, { "epoch": 0.12, "learning_rate": 0.0019552000653260816, "loss": 1.7627, "step": 1246 }, { "epoch": 0.12, "learning_rate": 0.0019551055163315205, "loss": 1.8184, "step": 1247 }, { "epoch": 0.12, "learning_rate": 0.0019550108699615777, "loss": 1.8906, "step": 1248 }, { "epoch": 0.12, "learning_rate": 0.0019549161262259027, "loss": 2.001, "step": 1249 }, { "epoch": 0.12, "learning_rate": 0.0019548212851341544, "loss": 1.8613, "step": 1250 }, { "epoch": 0.12, "learning_rate": 0.001954726346696002, "loss": 1.8564, "step": 1251 }, { "epoch": 0.12, "learning_rate": 0.0019546313109211253, "loss": 1.875, "step": 1252 }, { "epoch": 0.12, "learning_rate": 0.001954536177819213, "loss": 1.9404, "step": 1253 }, { "epoch": 0.12, "learning_rate": 0.001954440947399965, "loss": 1.9375, "step": 1254 }, { "epoch": 0.12, "learning_rate": 0.0019543456196730886, "loss": 1.8203, "step": 1255 }, { "epoch": 0.12, "learning_rate": 0.001954250194648304, "loss": 1.9658, "step": 1256 }, { "epoch": 0.12, "learning_rate": 0.00195415467233534, "loss": 1.9287, "step": 1257 }, { "epoch": 0.12, "learning_rate": 0.001954059052743934, "loss": 1.9961, "step": 1258 }, { "epoch": 0.12, "learning_rate": 0.0019539633358838368, "loss": 1.9082, "step": 1259 }, { "epoch": 0.12, "learning_rate": 0.001953867521764805, "loss": 1.8662, "step": 1260 }, { "epoch": 0.12, "learning_rate": 0.0019537716103966087, "loss": 1.9336, "step": 1261 }, { "epoch": 0.12, "learning_rate": 0.001953675601789025, "loss": 1.8574, "step": 1262 }, { "epoch": 0.12, "learning_rate": 0.0019535794959518425, "loss": 1.8428, "step": 1263 }, { "epoch": 0.12, "learning_rate": 0.00195348329289486, "loss": 1.8516, "step": 1264 }, { "epoch": 0.12, "learning_rate": 0.0019533869926278854, "loss": 1.7949, "step": 1265 }, { "epoch": 0.12, "learning_rate": 0.0019532905951607366, "loss": 1.8691, "step": 1266 }, { "epoch": 0.12, "learning_rate": 0.0019531941005032417, "loss": 1.8623, "step": 1267 }, { "epoch": 0.12, "learning_rate": 0.0019530975086652384, "loss": 1.9414, "step": 1268 }, { "epoch": 0.13, "learning_rate": 0.001953000819656575, "loss": 1.8418, "step": 1269 }, { "epoch": 0.13, "learning_rate": 0.0019529040334871081, "loss": 1.875, "step": 1270 }, { "epoch": 0.13, "learning_rate": 0.001952807150166706, "loss": 1.7822, "step": 1271 }, { "epoch": 0.13, "learning_rate": 0.0019527101697052467, "loss": 1.875, "step": 1272 }, { "epoch": 0.13, "learning_rate": 0.0019526130921126165, "loss": 1.9492, "step": 1273 }, { "epoch": 0.13, "learning_rate": 0.0019525159173987138, "loss": 1.9043, "step": 1274 }, { "epoch": 0.13, "learning_rate": 0.0019524186455734448, "loss": 1.9043, "step": 1275 }, { "epoch": 0.13, "learning_rate": 0.0019523212766467271, "loss": 1.9561, "step": 1276 }, { "epoch": 0.13, "learning_rate": 0.0019522238106284878, "loss": 1.8857, "step": 1277 }, { "epoch": 0.13, "learning_rate": 0.0019521262475286636, "loss": 1.8848, "step": 1278 }, { "epoch": 0.13, "learning_rate": 0.0019520285873572015, "loss": 1.9707, "step": 1279 }, { "epoch": 0.13, "learning_rate": 0.0019519308301240578, "loss": 1.8633, "step": 1280 }, { "epoch": 0.13, "learning_rate": 0.0019518329758391993, "loss": 1.8447, "step": 1281 }, { "epoch": 0.13, "learning_rate": 0.001951735024512603, "loss": 1.9775, "step": 1282 }, { "epoch": 0.13, "learning_rate": 0.0019516369761542543, "loss": 1.8232, "step": 1283 }, { "epoch": 0.13, "learning_rate": 0.0019515388307741503, "loss": 1.9336, "step": 1284 }, { "epoch": 0.13, "learning_rate": 0.0019514405883822967, "loss": 1.9717, "step": 1285 }, { "epoch": 0.13, "learning_rate": 0.0019513422489887099, "loss": 1.8799, "step": 1286 }, { "epoch": 0.13, "learning_rate": 0.0019512438126034156, "loss": 1.9043, "step": 1287 }, { "epoch": 0.13, "learning_rate": 0.0019511452792364496, "loss": 1.8418, "step": 1288 }, { "epoch": 0.13, "learning_rate": 0.0019510466488978579, "loss": 1.8809, "step": 1289 }, { "epoch": 0.13, "learning_rate": 0.0019509479215976957, "loss": 1.8115, "step": 1290 }, { "epoch": 0.13, "learning_rate": 0.001950849097346029, "loss": 1.8721, "step": 1291 }, { "epoch": 0.13, "learning_rate": 0.0019507501761529325, "loss": 1.8154, "step": 1292 }, { "epoch": 0.13, "learning_rate": 0.0019506511580284923, "loss": 1.9561, "step": 1293 }, { "epoch": 0.13, "learning_rate": 0.0019505520429828028, "loss": 1.8145, "step": 1294 }, { "epoch": 0.13, "learning_rate": 0.0019504528310259693, "loss": 1.9824, "step": 1295 }, { "epoch": 0.13, "learning_rate": 0.0019503535221681069, "loss": 1.7471, "step": 1296 }, { "epoch": 0.13, "learning_rate": 0.00195025411641934, "loss": 1.8594, "step": 1297 }, { "epoch": 0.13, "learning_rate": 0.0019501546137898039, "loss": 1.9336, "step": 1298 }, { "epoch": 0.13, "learning_rate": 0.0019500550142896424, "loss": 1.873, "step": 1299 }, { "epoch": 0.13, "learning_rate": 0.0019499553179290103, "loss": 1.9268, "step": 1300 }, { "epoch": 0.13, "learning_rate": 0.001949855524718072, "loss": 1.9023, "step": 1301 }, { "epoch": 0.13, "learning_rate": 0.0019497556346670013, "loss": 1.8145, "step": 1302 }, { "epoch": 0.13, "learning_rate": 0.0019496556477859826, "loss": 1.8896, "step": 1303 }, { "epoch": 0.13, "learning_rate": 0.0019495555640852095, "loss": 1.8408, "step": 1304 }, { "epoch": 0.13, "learning_rate": 0.0019494553835748862, "loss": 1.8965, "step": 1305 }, { "epoch": 0.13, "learning_rate": 0.0019493551062652256, "loss": 1.7998, "step": 1306 }, { "epoch": 0.13, "learning_rate": 0.001949254732166452, "loss": 1.9482, "step": 1307 }, { "epoch": 0.13, "learning_rate": 0.0019491542612887986, "loss": 1.8984, "step": 1308 }, { "epoch": 0.13, "learning_rate": 0.0019490536936425083, "loss": 1.8398, "step": 1309 }, { "epoch": 0.13, "learning_rate": 0.0019489530292378346, "loss": 1.9277, "step": 1310 }, { "epoch": 0.13, "learning_rate": 0.0019488522680850405, "loss": 1.9805, "step": 1311 }, { "epoch": 0.13, "learning_rate": 0.0019487514101943983, "loss": 1.8418, "step": 1312 }, { "epoch": 0.13, "learning_rate": 0.0019486504555761913, "loss": 1.9365, "step": 1313 }, { "epoch": 0.13, "learning_rate": 0.0019485494042407122, "loss": 1.8369, "step": 1314 }, { "epoch": 0.13, "learning_rate": 0.0019484482561982627, "loss": 1.9551, "step": 1315 }, { "epoch": 0.13, "learning_rate": 0.001948347011459156, "loss": 1.8184, "step": 1316 }, { "epoch": 0.13, "learning_rate": 0.0019482456700337131, "loss": 1.8633, "step": 1317 }, { "epoch": 0.13, "learning_rate": 0.0019481442319322674, "loss": 1.877, "step": 1318 }, { "epoch": 0.13, "learning_rate": 0.0019480426971651595, "loss": 1.9551, "step": 1319 }, { "epoch": 0.13, "learning_rate": 0.001947941065742742, "loss": 1.7705, "step": 1320 }, { "epoch": 0.13, "learning_rate": 0.001947839337675376, "loss": 1.8711, "step": 1321 }, { "epoch": 0.13, "learning_rate": 0.0019477375129734332, "loss": 1.833, "step": 1322 }, { "epoch": 0.13, "learning_rate": 0.0019476355916472948, "loss": 1.957, "step": 1323 }, { "epoch": 0.13, "learning_rate": 0.0019475335737073516, "loss": 1.8164, "step": 1324 }, { "epoch": 0.13, "learning_rate": 0.001947431459164005, "loss": 1.9766, "step": 1325 }, { "epoch": 0.13, "learning_rate": 0.001947329248027666, "loss": 1.9023, "step": 1326 }, { "epoch": 0.13, "learning_rate": 0.0019472269403087545, "loss": 1.9287, "step": 1327 }, { "epoch": 0.13, "learning_rate": 0.001947124536017702, "loss": 2.0, "step": 1328 }, { "epoch": 0.13, "learning_rate": 0.0019470220351649479, "loss": 1.7012, "step": 1329 }, { "epoch": 0.13, "learning_rate": 0.0019469194377609435, "loss": 1.9453, "step": 1330 }, { "epoch": 0.13, "learning_rate": 0.0019468167438161476, "loss": 1.8613, "step": 1331 }, { "epoch": 0.13, "learning_rate": 0.0019467139533410312, "loss": 1.793, "step": 1332 }, { "epoch": 0.13, "learning_rate": 0.0019466110663460734, "loss": 1.7627, "step": 1333 }, { "epoch": 0.13, "learning_rate": 0.0019465080828417643, "loss": 1.8691, "step": 1334 }, { "epoch": 0.13, "learning_rate": 0.0019464050028386026, "loss": 1.8945, "step": 1335 }, { "epoch": 0.13, "learning_rate": 0.0019463018263470985, "loss": 1.9111, "step": 1336 }, { "epoch": 0.13, "learning_rate": 0.0019461985533777702, "loss": 1.8545, "step": 1337 }, { "epoch": 0.13, "learning_rate": 0.001946095183941147, "loss": 1.8408, "step": 1338 }, { "epoch": 0.13, "learning_rate": 0.0019459917180477677, "loss": 1.9824, "step": 1339 }, { "epoch": 0.13, "learning_rate": 0.0019458881557081812, "loss": 1.8584, "step": 1340 }, { "epoch": 0.13, "learning_rate": 0.0019457844969329452, "loss": 1.8867, "step": 1341 }, { "epoch": 0.13, "learning_rate": 0.0019456807417326289, "loss": 1.9102, "step": 1342 }, { "epoch": 0.13, "learning_rate": 0.0019455768901178094, "loss": 1.9414, "step": 1343 }, { "epoch": 0.13, "learning_rate": 0.0019454729420990754, "loss": 1.8525, "step": 1344 }, { "epoch": 0.13, "learning_rate": 0.0019453688976870246, "loss": 1.7969, "step": 1345 }, { "epoch": 0.13, "learning_rate": 0.0019452647568922639, "loss": 1.8564, "step": 1346 }, { "epoch": 0.13, "learning_rate": 0.0019451605197254114, "loss": 1.916, "step": 1347 }, { "epoch": 0.13, "learning_rate": 0.0019450561861970943, "loss": 1.8408, "step": 1348 }, { "epoch": 0.13, "learning_rate": 0.0019449517563179493, "loss": 1.8594, "step": 1349 }, { "epoch": 0.13, "learning_rate": 0.0019448472300986235, "loss": 1.9189, "step": 1350 }, { "epoch": 0.13, "learning_rate": 0.0019447426075497737, "loss": 1.8301, "step": 1351 }, { "epoch": 0.13, "learning_rate": 0.001944637888682066, "loss": 1.8838, "step": 1352 }, { "epoch": 0.13, "learning_rate": 0.0019445330735061775, "loss": 1.9072, "step": 1353 }, { "epoch": 0.13, "learning_rate": 0.0019444281620327936, "loss": 1.918, "step": 1354 }, { "epoch": 0.13, "learning_rate": 0.0019443231542726107, "loss": 1.9014, "step": 1355 }, { "epoch": 0.13, "learning_rate": 0.0019442180502363346, "loss": 1.9688, "step": 1356 }, { "epoch": 0.13, "learning_rate": 0.0019441128499346808, "loss": 1.8506, "step": 1357 }, { "epoch": 0.13, "learning_rate": 0.0019440075533783748, "loss": 1.9736, "step": 1358 }, { "epoch": 0.13, "learning_rate": 0.0019439021605781516, "loss": 1.9336, "step": 1359 }, { "epoch": 0.13, "learning_rate": 0.0019437966715447569, "loss": 1.9033, "step": 1360 }, { "epoch": 0.13, "learning_rate": 0.0019436910862889447, "loss": 1.876, "step": 1361 }, { "epoch": 0.13, "learning_rate": 0.0019435854048214804, "loss": 1.8213, "step": 1362 }, { "epoch": 0.13, "learning_rate": 0.0019434796271531381, "loss": 1.8916, "step": 1363 }, { "epoch": 0.13, "learning_rate": 0.0019433737532947022, "loss": 1.9092, "step": 1364 }, { "epoch": 0.13, "learning_rate": 0.001943267783256967, "loss": 1.9014, "step": 1365 }, { "epoch": 0.13, "learning_rate": 0.0019431617170507362, "loss": 1.7812, "step": 1366 }, { "epoch": 0.13, "learning_rate": 0.0019430555546868231, "loss": 1.8418, "step": 1367 }, { "epoch": 0.13, "learning_rate": 0.001942949296176052, "loss": 1.8965, "step": 1368 }, { "epoch": 0.13, "learning_rate": 0.0019428429415292556, "loss": 1.8213, "step": 1369 }, { "epoch": 0.14, "learning_rate": 0.0019427364907572777, "loss": 1.8115, "step": 1370 }, { "epoch": 0.14, "learning_rate": 0.0019426299438709707, "loss": 1.8926, "step": 1371 }, { "epoch": 0.14, "learning_rate": 0.0019425233008811974, "loss": 1.8154, "step": 1372 }, { "epoch": 0.14, "learning_rate": 0.0019424165617988303, "loss": 1.7412, "step": 1373 }, { "epoch": 0.14, "learning_rate": 0.0019423097266347516, "loss": 1.7969, "step": 1374 }, { "epoch": 0.14, "learning_rate": 0.0019422027953998537, "loss": 1.8965, "step": 1375 }, { "epoch": 0.14, "learning_rate": 0.0019420957681050384, "loss": 1.8672, "step": 1376 }, { "epoch": 0.14, "learning_rate": 0.0019419886447612177, "loss": 1.8555, "step": 1377 }, { "epoch": 0.14, "learning_rate": 0.0019418814253793122, "loss": 1.916, "step": 1378 }, { "epoch": 0.14, "learning_rate": 0.0019417741099702539, "loss": 1.9062, "step": 1379 }, { "epoch": 0.14, "learning_rate": 0.0019416666985449839, "loss": 1.9248, "step": 1380 }, { "epoch": 0.14, "learning_rate": 0.0019415591911144529, "loss": 1.8467, "step": 1381 }, { "epoch": 0.14, "learning_rate": 0.0019414515876896212, "loss": 1.8506, "step": 1382 }, { "epoch": 0.14, "learning_rate": 0.00194134388828146, "loss": 1.9395, "step": 1383 }, { "epoch": 0.14, "learning_rate": 0.001941236092900949, "loss": 1.8184, "step": 1384 }, { "epoch": 0.14, "learning_rate": 0.001941128201559078, "loss": 1.9707, "step": 1385 }, { "epoch": 0.14, "learning_rate": 0.0019410202142668474, "loss": 1.8594, "step": 1386 }, { "epoch": 0.14, "learning_rate": 0.0019409121310352663, "loss": 1.875, "step": 1387 }, { "epoch": 0.14, "learning_rate": 0.0019408039518753544, "loss": 1.8799, "step": 1388 }, { "epoch": 0.14, "learning_rate": 0.0019406956767981404, "loss": 1.7881, "step": 1389 }, { "epoch": 0.14, "learning_rate": 0.0019405873058146637, "loss": 1.9268, "step": 1390 }, { "epoch": 0.14, "learning_rate": 0.0019404788389359726, "loss": 1.8047, "step": 1391 }, { "epoch": 0.14, "learning_rate": 0.0019403702761731259, "loss": 1.792, "step": 1392 }, { "epoch": 0.14, "learning_rate": 0.0019402616175371917, "loss": 1.8643, "step": 1393 }, { "epoch": 0.14, "learning_rate": 0.0019401528630392476, "loss": 1.8125, "step": 1394 }, { "epoch": 0.14, "learning_rate": 0.001940044012690382, "loss": 1.8428, "step": 1395 }, { "epoch": 0.14, "learning_rate": 0.0019399350665016924, "loss": 1.8936, "step": 1396 }, { "epoch": 0.14, "learning_rate": 0.0019398260244842859, "loss": 1.792, "step": 1397 }, { "epoch": 0.14, "learning_rate": 0.0019397168866492796, "loss": 1.8994, "step": 1398 }, { "epoch": 0.14, "learning_rate": 0.0019396076530078005, "loss": 1.8203, "step": 1399 }, { "epoch": 0.14, "learning_rate": 0.0019394983235709851, "loss": 1.8916, "step": 1400 }, { "epoch": 0.14, "learning_rate": 0.00193938889834998, "loss": 1.9736, "step": 1401 }, { "epoch": 0.14, "learning_rate": 0.0019392793773559411, "loss": 1.9492, "step": 1402 }, { "epoch": 0.14, "learning_rate": 0.001939169760600035, "loss": 1.7852, "step": 1403 }, { "epoch": 0.14, "learning_rate": 0.0019390600480934364, "loss": 1.8691, "step": 1404 }, { "epoch": 0.14, "learning_rate": 0.0019389502398473315, "loss": 1.8457, "step": 1405 }, { "epoch": 0.14, "learning_rate": 0.0019388403358729153, "loss": 1.9014, "step": 1406 }, { "epoch": 0.14, "learning_rate": 0.0019387303361813929, "loss": 1.8213, "step": 1407 }, { "epoch": 0.14, "learning_rate": 0.0019386202407839787, "loss": 1.9297, "step": 1408 }, { "epoch": 0.14, "learning_rate": 0.0019385100496918976, "loss": 1.9248, "step": 1409 }, { "epoch": 0.14, "learning_rate": 0.0019383997629163834, "loss": 1.9297, "step": 1410 }, { "epoch": 0.14, "learning_rate": 0.0019382893804686808, "loss": 1.8799, "step": 1411 }, { "epoch": 0.14, "learning_rate": 0.0019381789023600428, "loss": 1.8721, "step": 1412 }, { "epoch": 0.14, "learning_rate": 0.0019380683286017335, "loss": 1.749, "step": 1413 }, { "epoch": 0.14, "learning_rate": 0.0019379576592050258, "loss": 1.8301, "step": 1414 }, { "epoch": 0.14, "learning_rate": 0.001937846894181203, "loss": 1.8906, "step": 1415 }, { "epoch": 0.14, "learning_rate": 0.0019377360335415578, "loss": 1.9053, "step": 1416 }, { "epoch": 0.14, "learning_rate": 0.0019376250772973926, "loss": 1.8818, "step": 1417 }, { "epoch": 0.14, "learning_rate": 0.0019375140254600197, "loss": 1.8867, "step": 1418 }, { "epoch": 0.14, "learning_rate": 0.0019374028780407612, "loss": 1.9453, "step": 1419 }, { "epoch": 0.14, "learning_rate": 0.0019372916350509485, "loss": 1.8428, "step": 1420 }, { "epoch": 0.14, "learning_rate": 0.0019371802965019236, "loss": 1.8721, "step": 1421 }, { "epoch": 0.14, "learning_rate": 0.0019370688624050375, "loss": 1.8965, "step": 1422 }, { "epoch": 0.14, "learning_rate": 0.0019369573327716512, "loss": 1.8428, "step": 1423 }, { "epoch": 0.14, "learning_rate": 0.0019368457076131354, "loss": 1.8066, "step": 1424 }, { "epoch": 0.14, "learning_rate": 0.0019367339869408708, "loss": 1.8828, "step": 1425 }, { "epoch": 0.14, "learning_rate": 0.001936622170766247, "loss": 1.8984, "step": 1426 }, { "epoch": 0.14, "learning_rate": 0.0019365102591006648, "loss": 1.7188, "step": 1427 }, { "epoch": 0.14, "learning_rate": 0.001936398251955533, "loss": 1.9365, "step": 1428 }, { "epoch": 0.14, "learning_rate": 0.0019362861493422714, "loss": 1.7021, "step": 1429 }, { "epoch": 0.14, "learning_rate": 0.0019361739512723096, "loss": 1.7354, "step": 1430 }, { "epoch": 0.14, "learning_rate": 0.0019360616577570858, "loss": 1.8525, "step": 1431 }, { "epoch": 0.14, "learning_rate": 0.0019359492688080487, "loss": 1.9648, "step": 1432 }, { "epoch": 0.14, "learning_rate": 0.0019358367844366567, "loss": 1.8369, "step": 1433 }, { "epoch": 0.14, "learning_rate": 0.001935724204654378, "loss": 1.7695, "step": 1434 }, { "epoch": 0.14, "learning_rate": 0.0019356115294726906, "loss": 1.8369, "step": 1435 }, { "epoch": 0.14, "learning_rate": 0.0019354987589030816, "loss": 1.834, "step": 1436 }, { "epoch": 0.14, "learning_rate": 0.0019353858929570481, "loss": 1.7861, "step": 1437 }, { "epoch": 0.14, "learning_rate": 0.0019352729316460975, "loss": 1.8008, "step": 1438 }, { "epoch": 0.14, "learning_rate": 0.0019351598749817462, "loss": 1.9639, "step": 1439 }, { "epoch": 0.14, "learning_rate": 0.0019350467229755212, "loss": 1.8242, "step": 1440 }, { "epoch": 0.14, "learning_rate": 0.0019349334756389577, "loss": 1.8662, "step": 1441 }, { "epoch": 0.14, "learning_rate": 0.001934820132983602, "loss": 1.9385, "step": 1442 }, { "epoch": 0.14, "learning_rate": 0.00193470669502101, "loss": 1.8516, "step": 1443 }, { "epoch": 0.14, "learning_rate": 0.0019345931617627463, "loss": 1.8135, "step": 1444 }, { "epoch": 0.14, "learning_rate": 0.0019344795332203863, "loss": 1.7998, "step": 1445 }, { "epoch": 0.14, "learning_rate": 0.0019343658094055148, "loss": 1.8477, "step": 1446 }, { "epoch": 0.14, "learning_rate": 0.0019342519903297259, "loss": 1.834, "step": 1447 }, { "epoch": 0.14, "learning_rate": 0.001934138076004624, "loss": 1.9102, "step": 1448 }, { "epoch": 0.14, "learning_rate": 0.0019340240664418229, "loss": 1.7979, "step": 1449 }, { "epoch": 0.14, "learning_rate": 0.001933909961652946, "loss": 1.9023, "step": 1450 }, { "epoch": 0.14, "learning_rate": 0.001933795761649627, "loss": 1.8516, "step": 1451 }, { "epoch": 0.14, "learning_rate": 0.0019336814664435082, "loss": 1.9297, "step": 1452 }, { "epoch": 0.14, "learning_rate": 0.001933567076046243, "loss": 1.8232, "step": 1453 }, { "epoch": 0.14, "learning_rate": 0.0019334525904694934, "loss": 1.8633, "step": 1454 }, { "epoch": 0.14, "learning_rate": 0.0019333380097249313, "loss": 1.8232, "step": 1455 }, { "epoch": 0.14, "learning_rate": 0.0019332233338242388, "loss": 1.8984, "step": 1456 }, { "epoch": 0.14, "learning_rate": 0.0019331085627791074, "loss": 1.7754, "step": 1457 }, { "epoch": 0.14, "learning_rate": 0.0019329936966012383, "loss": 1.8164, "step": 1458 }, { "epoch": 0.14, "learning_rate": 0.001932878735302342, "loss": 1.8477, "step": 1459 }, { "epoch": 0.14, "learning_rate": 0.0019327636788941396, "loss": 1.9004, "step": 1460 }, { "epoch": 0.14, "learning_rate": 0.001932648527388361, "loss": 1.9082, "step": 1461 }, { "epoch": 0.14, "learning_rate": 0.0019325332807967467, "loss": 1.8887, "step": 1462 }, { "epoch": 0.14, "learning_rate": 0.0019324179391310454, "loss": 1.8076, "step": 1463 }, { "epoch": 0.14, "learning_rate": 0.0019323025024030177, "loss": 1.9141, "step": 1464 }, { "epoch": 0.14, "learning_rate": 0.0019321869706244318, "loss": 1.8643, "step": 1465 }, { "epoch": 0.14, "learning_rate": 0.0019320713438070666, "loss": 1.8457, "step": 1466 }, { "epoch": 0.14, "learning_rate": 0.0019319556219627111, "loss": 1.8613, "step": 1467 }, { "epoch": 0.14, "learning_rate": 0.0019318398051031624, "loss": 1.8877, "step": 1468 }, { "epoch": 0.14, "learning_rate": 0.0019317238932402292, "loss": 1.8438, "step": 1469 }, { "epoch": 0.14, "learning_rate": 0.001931607886385729, "loss": 1.9072, "step": 1470 }, { "epoch": 0.15, "learning_rate": 0.001931491784551488, "loss": 1.8145, "step": 1471 }, { "epoch": 0.15, "learning_rate": 0.0019313755877493444, "loss": 1.8564, "step": 1472 }, { "epoch": 0.15, "learning_rate": 0.001931259295991144, "loss": 1.9424, "step": 1473 }, { "epoch": 0.15, "learning_rate": 0.0019311429092887429, "loss": 1.7422, "step": 1474 }, { "epoch": 0.15, "learning_rate": 0.0019310264276540074, "loss": 1.832, "step": 1475 }, { "epoch": 0.15, "learning_rate": 0.001930909851098813, "loss": 1.9209, "step": 1476 }, { "epoch": 0.15, "learning_rate": 0.0019307931796350448, "loss": 1.9102, "step": 1477 }, { "epoch": 0.15, "learning_rate": 0.001930676413274598, "loss": 1.9727, "step": 1478 }, { "epoch": 0.15, "learning_rate": 0.001930559552029377, "loss": 1.7441, "step": 1479 }, { "epoch": 0.15, "learning_rate": 0.001930442595911296, "loss": 1.8379, "step": 1480 }, { "epoch": 0.15, "learning_rate": 0.0019303255449322797, "loss": 1.9404, "step": 1481 }, { "epoch": 0.15, "learning_rate": 0.0019302083991042608, "loss": 1.7637, "step": 1482 }, { "epoch": 0.15, "learning_rate": 0.001930091158439183, "loss": 1.8711, "step": 1483 }, { "epoch": 0.15, "learning_rate": 0.0019299738229489993, "loss": 1.918, "step": 1484 }, { "epoch": 0.15, "learning_rate": 0.0019298563926456725, "loss": 1.8789, "step": 1485 }, { "epoch": 0.15, "learning_rate": 0.0019297388675411744, "loss": 1.8018, "step": 1486 }, { "epoch": 0.15, "learning_rate": 0.0019296212476474877, "loss": 1.8135, "step": 1487 }, { "epoch": 0.15, "learning_rate": 0.0019295035329766033, "loss": 1.8164, "step": 1488 }, { "epoch": 0.15, "learning_rate": 0.001929385723540523, "loss": 1.8145, "step": 1489 }, { "epoch": 0.15, "learning_rate": 0.0019292678193512578, "loss": 1.9297, "step": 1490 }, { "epoch": 0.15, "learning_rate": 0.001929149820420828, "loss": 1.8623, "step": 1491 }, { "epoch": 0.15, "learning_rate": 0.0019290317267612641, "loss": 1.8486, "step": 1492 }, { "epoch": 0.15, "learning_rate": 0.001928913538384606, "loss": 1.915, "step": 1493 }, { "epoch": 0.15, "learning_rate": 0.0019287952553029028, "loss": 1.7705, "step": 1494 }, { "epoch": 0.15, "learning_rate": 0.0019286768775282149, "loss": 1.8789, "step": 1495 }, { "epoch": 0.15, "learning_rate": 0.00192855840507261, "loss": 1.8682, "step": 1496 }, { "epoch": 0.15, "learning_rate": 0.0019284398379481673, "loss": 1.8301, "step": 1497 }, { "epoch": 0.15, "learning_rate": 0.0019283211761669753, "loss": 1.7617, "step": 1498 }, { "epoch": 0.15, "learning_rate": 0.001928202419741131, "loss": 1.7109, "step": 1499 }, { "epoch": 0.15, "learning_rate": 0.0019280835686827427, "loss": 1.8857, "step": 1500 }, { "epoch": 0.15, "learning_rate": 0.001927964623003927, "loss": 1.8193, "step": 1501 }, { "epoch": 0.15, "learning_rate": 0.0019278455827168109, "loss": 1.7949, "step": 1502 }, { "epoch": 0.15, "learning_rate": 0.001927726447833531, "loss": 1.8877, "step": 1503 }, { "epoch": 0.15, "learning_rate": 0.0019276072183662338, "loss": 1.9717, "step": 1504 }, { "epoch": 0.15, "learning_rate": 0.0019274878943270738, "loss": 1.7959, "step": 1505 }, { "epoch": 0.15, "learning_rate": 0.0019273684757282175, "loss": 1.8496, "step": 1506 }, { "epoch": 0.15, "learning_rate": 0.0019272489625818394, "loss": 1.8975, "step": 1507 }, { "epoch": 0.15, "learning_rate": 0.0019271293549001243, "loss": 1.8271, "step": 1508 }, { "epoch": 0.15, "learning_rate": 0.0019270096526952667, "loss": 1.8174, "step": 1509 }, { "epoch": 0.15, "learning_rate": 0.0019268898559794702, "loss": 1.8037, "step": 1510 }, { "epoch": 0.15, "learning_rate": 0.0019267699647649485, "loss": 1.9033, "step": 1511 }, { "epoch": 0.15, "learning_rate": 0.001926649979063925, "loss": 1.8125, "step": 1512 }, { "epoch": 0.15, "learning_rate": 0.0019265298988886323, "loss": 1.792, "step": 1513 }, { "epoch": 0.15, "learning_rate": 0.001926409724251313, "loss": 1.8506, "step": 1514 }, { "epoch": 0.15, "learning_rate": 0.001926289455164219, "loss": 1.8037, "step": 1515 }, { "epoch": 0.15, "learning_rate": 0.0019261690916396123, "loss": 1.7891, "step": 1516 }, { "epoch": 0.15, "learning_rate": 0.001926048633689764, "loss": 1.8096, "step": 1517 }, { "epoch": 0.15, "learning_rate": 0.0019259280813269553, "loss": 1.8457, "step": 1518 }, { "epoch": 0.15, "learning_rate": 0.0019258074345634769, "loss": 1.7842, "step": 1519 }, { "epoch": 0.15, "learning_rate": 0.001925686693411629, "loss": 1.8242, "step": 1520 }, { "epoch": 0.15, "learning_rate": 0.0019255658578837208, "loss": 1.9121, "step": 1521 }, { "epoch": 0.15, "learning_rate": 0.0019254449279920724, "loss": 1.8779, "step": 1522 }, { "epoch": 0.15, "learning_rate": 0.0019253239037490132, "loss": 1.8252, "step": 1523 }, { "epoch": 0.15, "learning_rate": 0.0019252027851668812, "loss": 1.8037, "step": 1524 }, { "epoch": 0.15, "learning_rate": 0.0019250815722580252, "loss": 1.8447, "step": 1525 }, { "epoch": 0.15, "learning_rate": 0.001924960265034803, "loss": 1.7236, "step": 1526 }, { "epoch": 0.15, "learning_rate": 0.0019248388635095823, "loss": 1.875, "step": 1527 }, { "epoch": 0.15, "learning_rate": 0.0019247173676947402, "loss": 1.9316, "step": 1528 }, { "epoch": 0.15, "learning_rate": 0.0019245957776026638, "loss": 1.9316, "step": 1529 }, { "epoch": 0.15, "learning_rate": 0.0019244740932457488, "loss": 1.8711, "step": 1530 }, { "epoch": 0.15, "learning_rate": 0.0019243523146364016, "loss": 1.793, "step": 1531 }, { "epoch": 0.15, "learning_rate": 0.0019242304417870382, "loss": 1.8447, "step": 1532 }, { "epoch": 0.15, "learning_rate": 0.0019241084747100832, "loss": 1.917, "step": 1533 }, { "epoch": 0.15, "learning_rate": 0.001923986413417972, "loss": 1.9229, "step": 1534 }, { "epoch": 0.15, "learning_rate": 0.0019238642579231487, "loss": 1.8154, "step": 1535 }, { "epoch": 0.15, "learning_rate": 0.0019237420082380675, "loss": 1.6514, "step": 1536 }, { "epoch": 0.15, "learning_rate": 0.001923619664375192, "loss": 1.9346, "step": 1537 }, { "epoch": 0.15, "learning_rate": 0.0019234972263469953, "loss": 1.833, "step": 1538 }, { "epoch": 0.15, "learning_rate": 0.0019233746941659607, "loss": 1.7959, "step": 1539 }, { "epoch": 0.15, "learning_rate": 0.0019232520678445803, "loss": 1.8193, "step": 1540 }, { "epoch": 0.15, "learning_rate": 0.0019231293473953564, "loss": 1.791, "step": 1541 }, { "epoch": 0.15, "learning_rate": 0.0019230065328308004, "loss": 1.8799, "step": 1542 }, { "epoch": 0.15, "learning_rate": 0.0019228836241634336, "loss": 1.8975, "step": 1543 }, { "epoch": 0.15, "learning_rate": 0.0019227606214057872, "loss": 1.8936, "step": 1544 }, { "epoch": 0.15, "learning_rate": 0.0019226375245704008, "loss": 1.832, "step": 1545 }, { "epoch": 0.15, "learning_rate": 0.0019225143336698257, "loss": 1.7705, "step": 1546 }, { "epoch": 0.15, "learning_rate": 0.0019223910487166202, "loss": 1.7764, "step": 1547 }, { "epoch": 0.15, "learning_rate": 0.0019222676697233543, "loss": 1.8926, "step": 1548 }, { "epoch": 0.15, "learning_rate": 0.001922144196702607, "loss": 1.8906, "step": 1549 }, { "epoch": 0.15, "learning_rate": 0.0019220206296669657, "loss": 1.8545, "step": 1550 }, { "epoch": 0.15, "learning_rate": 0.0019218969686290294, "loss": 2.0098, "step": 1551 }, { "epoch": 0.15, "learning_rate": 0.001921773213601405, "loss": 1.9365, "step": 1552 }, { "epoch": 0.15, "learning_rate": 0.0019216493645967097, "loss": 1.8115, "step": 1553 }, { "epoch": 0.15, "learning_rate": 0.0019215254216275704, "loss": 1.8447, "step": 1554 }, { "epoch": 0.15, "learning_rate": 0.0019214013847066234, "loss": 1.8418, "step": 1555 }, { "epoch": 0.15, "learning_rate": 0.0019212772538465144, "loss": 1.8604, "step": 1556 }, { "epoch": 0.15, "learning_rate": 0.0019211530290598993, "loss": 1.8242, "step": 1557 }, { "epoch": 0.15, "learning_rate": 0.0019210287103594427, "loss": 1.8203, "step": 1558 }, { "epoch": 0.15, "learning_rate": 0.001920904297757819, "loss": 1.9355, "step": 1559 }, { "epoch": 0.15, "learning_rate": 0.0019207797912677127, "loss": 1.8145, "step": 1560 }, { "epoch": 0.15, "learning_rate": 0.0019206551909018176, "loss": 1.7871, "step": 1561 }, { "epoch": 0.15, "learning_rate": 0.001920530496672837, "loss": 1.6689, "step": 1562 }, { "epoch": 0.15, "learning_rate": 0.0019204057085934836, "loss": 1.9355, "step": 1563 }, { "epoch": 0.15, "learning_rate": 0.0019202808266764802, "loss": 1.9414, "step": 1564 }, { "epoch": 0.15, "learning_rate": 0.0019201558509345585, "loss": 1.874, "step": 1565 }, { "epoch": 0.15, "learning_rate": 0.00192003078138046, "loss": 1.8711, "step": 1566 }, { "epoch": 0.15, "learning_rate": 0.0019199056180269362, "loss": 1.8809, "step": 1567 }, { "epoch": 0.15, "learning_rate": 0.0019197803608867474, "loss": 1.8584, "step": 1568 }, { "epoch": 0.15, "learning_rate": 0.0019196550099726646, "loss": 1.8369, "step": 1569 }, { "epoch": 0.15, "learning_rate": 0.0019195295652974667, "loss": 1.915, "step": 1570 }, { "epoch": 0.15, "learning_rate": 0.0019194040268739436, "loss": 1.7119, "step": 1571 }, { "epoch": 0.15, "learning_rate": 0.0019192783947148944, "loss": 1.8506, "step": 1572 }, { "epoch": 0.16, "learning_rate": 0.0019191526688331275, "loss": 1.792, "step": 1573 }, { "epoch": 0.16, "learning_rate": 0.0019190268492414607, "loss": 1.9248, "step": 1574 }, { "epoch": 0.16, "learning_rate": 0.0019189009359527219, "loss": 1.999, "step": 1575 }, { "epoch": 0.16, "learning_rate": 0.001918774928979748, "loss": 1.8359, "step": 1576 }, { "epoch": 0.16, "learning_rate": 0.0019186488283353862, "loss": 1.7969, "step": 1577 }, { "epoch": 0.16, "learning_rate": 0.001918522634032492, "loss": 1.8604, "step": 1578 }, { "epoch": 0.16, "learning_rate": 0.0019183963460839323, "loss": 1.7236, "step": 1579 }, { "epoch": 0.16, "learning_rate": 0.0019182699645025813, "loss": 1.835, "step": 1580 }, { "epoch": 0.16, "learning_rate": 0.001918143489301325, "loss": 1.8838, "step": 1581 }, { "epoch": 0.16, "learning_rate": 0.001918016920493057, "loss": 1.8193, "step": 1582 }, { "epoch": 0.16, "learning_rate": 0.0019178902580906819, "loss": 1.8809, "step": 1583 }, { "epoch": 0.16, "learning_rate": 0.0019177635021071127, "loss": 1.8711, "step": 1584 }, { "epoch": 0.16, "learning_rate": 0.0019176366525552732, "loss": 1.8037, "step": 1585 }, { "epoch": 0.16, "learning_rate": 0.0019175097094480953, "loss": 1.8115, "step": 1586 }, { "epoch": 0.16, "learning_rate": 0.0019173826727985216, "loss": 1.7539, "step": 1587 }, { "epoch": 0.16, "learning_rate": 0.0019172555426195035, "loss": 1.8086, "step": 1588 }, { "epoch": 0.16, "learning_rate": 0.0019171283189240028, "loss": 1.9375, "step": 1589 }, { "epoch": 0.16, "learning_rate": 0.0019170010017249895, "loss": 1.8252, "step": 1590 }, { "epoch": 0.16, "learning_rate": 0.0019168735910354445, "loss": 1.832, "step": 1591 }, { "epoch": 0.16, "learning_rate": 0.0019167460868683573, "loss": 1.9033, "step": 1592 }, { "epoch": 0.16, "learning_rate": 0.0019166184892367274, "loss": 1.7803, "step": 1593 }, { "epoch": 0.16, "learning_rate": 0.001916490798153564, "loss": 1.8486, "step": 1594 }, { "epoch": 0.16, "learning_rate": 0.001916363013631885, "loss": 1.8623, "step": 1595 }, { "epoch": 0.16, "learning_rate": 0.0019162351356847185, "loss": 1.8682, "step": 1596 }, { "epoch": 0.16, "learning_rate": 0.0019161071643251021, "loss": 1.8008, "step": 1597 }, { "epoch": 0.16, "learning_rate": 0.0019159790995660828, "loss": 1.8809, "step": 1598 }, { "epoch": 0.16, "learning_rate": 0.0019158509414207172, "loss": 1.8594, "step": 1599 }, { "epoch": 0.16, "learning_rate": 0.0019157226899020707, "loss": 1.8223, "step": 1600 }, { "epoch": 0.16, "learning_rate": 0.00191559434502322, "loss": 1.834, "step": 1601 }, { "epoch": 0.16, "learning_rate": 0.0019154659067972491, "loss": 1.7559, "step": 1602 }, { "epoch": 0.16, "learning_rate": 0.0019153373752372534, "loss": 1.7998, "step": 1603 }, { "epoch": 0.16, "learning_rate": 0.0019152087503563364, "loss": 1.8027, "step": 1604 }, { "epoch": 0.16, "learning_rate": 0.0019150800321676119, "loss": 1.8164, "step": 1605 }, { "epoch": 0.16, "learning_rate": 0.0019149512206842031, "loss": 1.8105, "step": 1606 }, { "epoch": 0.16, "learning_rate": 0.001914822315919243, "loss": 1.7549, "step": 1607 }, { "epoch": 0.16, "learning_rate": 0.0019146933178858735, "loss": 1.8389, "step": 1608 }, { "epoch": 0.16, "learning_rate": 0.0019145642265972459, "loss": 1.9053, "step": 1609 }, { "epoch": 0.16, "learning_rate": 0.0019144350420665216, "loss": 1.832, "step": 1610 }, { "epoch": 0.16, "learning_rate": 0.0019143057643068715, "loss": 1.8809, "step": 1611 }, { "epoch": 0.16, "learning_rate": 0.0019141763933314758, "loss": 1.8945, "step": 1612 }, { "epoch": 0.16, "learning_rate": 0.0019140469291535238, "loss": 1.7666, "step": 1613 }, { "epoch": 0.16, "learning_rate": 0.001913917371786215, "loss": 1.9189, "step": 1614 }, { "epoch": 0.16, "learning_rate": 0.0019137877212427581, "loss": 1.8828, "step": 1615 }, { "epoch": 0.16, "learning_rate": 0.0019136579775363711, "loss": 1.8877, "step": 1616 }, { "epoch": 0.16, "learning_rate": 0.001913528140680282, "loss": 1.8057, "step": 1617 }, { "epoch": 0.16, "learning_rate": 0.0019133982106877276, "loss": 1.7656, "step": 1618 }, { "epoch": 0.16, "learning_rate": 0.001913268187571955, "loss": 1.8232, "step": 1619 }, { "epoch": 0.16, "learning_rate": 0.0019131380713462204, "loss": 1.7871, "step": 1620 }, { "epoch": 0.16, "learning_rate": 0.0019130078620237887, "loss": 1.8896, "step": 1621 }, { "epoch": 0.16, "learning_rate": 0.0019128775596179359, "loss": 1.8428, "step": 1622 }, { "epoch": 0.16, "learning_rate": 0.0019127471641419468, "loss": 1.8652, "step": 1623 }, { "epoch": 0.16, "learning_rate": 0.0019126166756091142, "loss": 1.8008, "step": 1624 }, { "epoch": 0.16, "learning_rate": 0.0019124860940327433, "loss": 1.8096, "step": 1625 }, { "epoch": 0.16, "learning_rate": 0.0019123554194261465, "loss": 1.8311, "step": 1626 }, { "epoch": 0.16, "learning_rate": 0.0019122246518026463, "loss": 1.8623, "step": 1627 }, { "epoch": 0.16, "learning_rate": 0.001912093791175575, "loss": 1.8232, "step": 1628 }, { "epoch": 0.16, "learning_rate": 0.0019119628375582744, "loss": 1.8105, "step": 1629 }, { "epoch": 0.16, "learning_rate": 0.0019118317909640947, "loss": 1.8906, "step": 1630 }, { "epoch": 0.16, "learning_rate": 0.0019117006514063975, "loss": 1.7734, "step": 1631 }, { "epoch": 0.16, "learning_rate": 0.001911569418898552, "loss": 1.7568, "step": 1632 }, { "epoch": 0.16, "learning_rate": 0.0019114380934539378, "loss": 1.8535, "step": 1633 }, { "epoch": 0.16, "learning_rate": 0.0019113066750859444, "loss": 1.8926, "step": 1634 }, { "epoch": 0.16, "learning_rate": 0.0019111751638079698, "loss": 1.7881, "step": 1635 }, { "epoch": 0.16, "learning_rate": 0.001911043559633422, "loss": 1.749, "step": 1636 }, { "epoch": 0.16, "learning_rate": 0.001910911862575718, "loss": 1.9307, "step": 1637 }, { "epoch": 0.16, "learning_rate": 0.0019107800726482851, "loss": 1.8975, "step": 1638 }, { "epoch": 0.16, "learning_rate": 0.0019106481898645599, "loss": 1.7021, "step": 1639 }, { "epoch": 0.16, "learning_rate": 0.0019105162142379873, "loss": 1.957, "step": 1640 }, { "epoch": 0.16, "learning_rate": 0.001910384145782023, "loss": 1.9258, "step": 1641 }, { "epoch": 0.16, "learning_rate": 0.0019102519845101318, "loss": 1.8135, "step": 1642 }, { "epoch": 0.16, "learning_rate": 0.0019101197304357879, "loss": 1.9082, "step": 1643 }, { "epoch": 0.16, "learning_rate": 0.0019099873835724747, "loss": 1.8584, "step": 1644 }, { "epoch": 0.16, "learning_rate": 0.0019098549439336857, "loss": 1.8486, "step": 1645 }, { "epoch": 0.16, "learning_rate": 0.0019097224115329227, "loss": 1.8252, "step": 1646 }, { "epoch": 0.16, "learning_rate": 0.0019095897863836981, "loss": 1.8232, "step": 1647 }, { "epoch": 0.16, "learning_rate": 0.0019094570684995338, "loss": 1.8379, "step": 1648 }, { "epoch": 0.16, "learning_rate": 0.0019093242578939603, "loss": 1.7998, "step": 1649 }, { "epoch": 0.16, "learning_rate": 0.0019091913545805181, "loss": 1.8154, "step": 1650 }, { "epoch": 0.16, "learning_rate": 0.0019090583585727567, "loss": 1.8496, "step": 1651 }, { "epoch": 0.16, "learning_rate": 0.001908925269884236, "loss": 1.8867, "step": 1652 }, { "epoch": 0.16, "learning_rate": 0.0019087920885285238, "loss": 1.8242, "step": 1653 }, { "epoch": 0.16, "learning_rate": 0.001908658814519199, "loss": 1.8594, "step": 1654 }, { "epoch": 0.16, "learning_rate": 0.001908525447869849, "loss": 1.834, "step": 1655 }, { "epoch": 0.16, "learning_rate": 0.0019083919885940711, "loss": 1.8428, "step": 1656 }, { "epoch": 0.16, "learning_rate": 0.0019082584367054714, "loss": 1.7148, "step": 1657 }, { "epoch": 0.16, "learning_rate": 0.001908124792217666, "loss": 1.8164, "step": 1658 }, { "epoch": 0.16, "learning_rate": 0.0019079910551442806, "loss": 1.7959, "step": 1659 }, { "epoch": 0.16, "learning_rate": 0.0019078572254989494, "loss": 1.8125, "step": 1660 }, { "epoch": 0.16, "learning_rate": 0.001907723303295317, "loss": 1.7988, "step": 1661 }, { "epoch": 0.16, "learning_rate": 0.0019075892885470375, "loss": 1.7637, "step": 1662 }, { "epoch": 0.16, "learning_rate": 0.0019074551812677733, "loss": 1.8369, "step": 1663 }, { "epoch": 0.16, "learning_rate": 0.0019073209814711976, "loss": 1.8164, "step": 1664 }, { "epoch": 0.16, "learning_rate": 0.0019071866891709921, "loss": 1.9404, "step": 1665 }, { "epoch": 0.16, "learning_rate": 0.0019070523043808482, "loss": 1.8369, "step": 1666 }, { "epoch": 0.16, "learning_rate": 0.0019069178271144667, "loss": 1.792, "step": 1667 }, { "epoch": 0.16, "learning_rate": 0.0019067832573855582, "loss": 1.7842, "step": 1668 }, { "epoch": 0.16, "learning_rate": 0.0019066485952078422, "loss": 1.6953, "step": 1669 }, { "epoch": 0.16, "learning_rate": 0.0019065138405950476, "loss": 1.876, "step": 1670 }, { "epoch": 0.16, "learning_rate": 0.0019063789935609137, "loss": 1.9121, "step": 1671 }, { "epoch": 0.16, "learning_rate": 0.0019062440541191877, "loss": 1.8584, "step": 1672 }, { "epoch": 0.16, "learning_rate": 0.0019061090222836274, "loss": 1.8438, "step": 1673 }, { "epoch": 0.17, "learning_rate": 0.0019059738980679998, "loss": 1.9473, "step": 1674 }, { "epoch": 0.17, "learning_rate": 0.0019058386814860807, "loss": 1.8232, "step": 1675 }, { "epoch": 0.17, "learning_rate": 0.001905703372551656, "loss": 1.7822, "step": 1676 }, { "epoch": 0.17, "learning_rate": 0.001905567971278521, "loss": 1.7637, "step": 1677 }, { "epoch": 0.17, "learning_rate": 0.00190543247768048, "loss": 1.8916, "step": 1678 }, { "epoch": 0.17, "learning_rate": 0.0019052968917713466, "loss": 1.7842, "step": 1679 }, { "epoch": 0.17, "learning_rate": 0.0019051612135649446, "loss": 1.8428, "step": 1680 }, { "epoch": 0.17, "learning_rate": 0.0019050254430751062, "loss": 1.9473, "step": 1681 }, { "epoch": 0.17, "learning_rate": 0.001904889580315674, "loss": 1.7441, "step": 1682 }, { "epoch": 0.17, "learning_rate": 0.0019047536253004999, "loss": 1.8145, "step": 1683 }, { "epoch": 0.17, "learning_rate": 0.0019046175780434438, "loss": 1.7842, "step": 1684 }, { "epoch": 0.17, "learning_rate": 0.0019044814385583769, "loss": 1.8867, "step": 1685 }, { "epoch": 0.17, "learning_rate": 0.0019043452068591787, "loss": 1.9502, "step": 1686 }, { "epoch": 0.17, "learning_rate": 0.0019042088829597384, "loss": 1.8408, "step": 1687 }, { "epoch": 0.17, "learning_rate": 0.0019040724668739544, "loss": 1.7998, "step": 1688 }, { "epoch": 0.17, "learning_rate": 0.0019039359586157346, "loss": 1.8799, "step": 1689 }, { "epoch": 0.17, "learning_rate": 0.0019037993581989968, "loss": 1.6768, "step": 1690 }, { "epoch": 0.17, "learning_rate": 0.0019036626656376674, "loss": 1.8145, "step": 1691 }, { "epoch": 0.17, "learning_rate": 0.0019035258809456826, "loss": 1.8135, "step": 1692 }, { "epoch": 0.17, "learning_rate": 0.001903389004136988, "loss": 1.8965, "step": 1693 }, { "epoch": 0.17, "learning_rate": 0.0019032520352255382, "loss": 1.8301, "step": 1694 }, { "epoch": 0.17, "learning_rate": 0.0019031149742252983, "loss": 1.8262, "step": 1695 }, { "epoch": 0.17, "learning_rate": 0.001902977821150241, "loss": 1.8232, "step": 1696 }, { "epoch": 0.17, "learning_rate": 0.0019028405760143504, "loss": 1.7539, "step": 1697 }, { "epoch": 0.17, "learning_rate": 0.0019027032388316183, "loss": 1.8242, "step": 1698 }, { "epoch": 0.17, "learning_rate": 0.0019025658096160467, "loss": 1.834, "step": 1699 }, { "epoch": 0.17, "learning_rate": 0.001902428288381647, "loss": 1.8066, "step": 1700 }, { "epoch": 0.17, "learning_rate": 0.0019022906751424398, "loss": 1.9248, "step": 1701 }, { "epoch": 0.17, "learning_rate": 0.001902152969912455, "loss": 1.75, "step": 1702 }, { "epoch": 0.17, "learning_rate": 0.0019020151727057323, "loss": 1.8311, "step": 1703 }, { "epoch": 0.17, "learning_rate": 0.0019018772835363198, "loss": 1.8496, "step": 1704 }, { "epoch": 0.17, "learning_rate": 0.0019017393024182763, "loss": 1.7695, "step": 1705 }, { "epoch": 0.17, "learning_rate": 0.001901601229365669, "loss": 1.7363, "step": 1706 }, { "epoch": 0.17, "learning_rate": 0.001901463064392575, "loss": 1.7891, "step": 1707 }, { "epoch": 0.17, "learning_rate": 0.0019013248075130803, "loss": 1.8281, "step": 1708 }, { "epoch": 0.17, "learning_rate": 0.0019011864587412809, "loss": 1.6816, "step": 1709 }, { "epoch": 0.17, "learning_rate": 0.0019010480180912813, "loss": 1.9014, "step": 1710 }, { "epoch": 0.17, "learning_rate": 0.0019009094855771962, "loss": 1.8252, "step": 1711 }, { "epoch": 0.17, "learning_rate": 0.0019007708612131494, "loss": 1.8662, "step": 1712 }, { "epoch": 0.17, "learning_rate": 0.0019006321450132737, "loss": 1.8447, "step": 1713 }, { "epoch": 0.17, "learning_rate": 0.001900493336991712, "loss": 1.7959, "step": 1714 }, { "epoch": 0.17, "learning_rate": 0.0019003544371626155, "loss": 1.876, "step": 1715 }, { "epoch": 0.17, "learning_rate": 0.0019002154455401458, "loss": 1.834, "step": 1716 }, { "epoch": 0.17, "learning_rate": 0.0019000763621384738, "loss": 1.8057, "step": 1717 }, { "epoch": 0.17, "learning_rate": 0.0018999371869717788, "loss": 1.9746, "step": 1718 }, { "epoch": 0.17, "learning_rate": 0.0018997979200542502, "loss": 1.7227, "step": 1719 }, { "epoch": 0.17, "learning_rate": 0.0018996585614000866, "loss": 1.8906, "step": 1720 }, { "epoch": 0.17, "learning_rate": 0.0018995191110234959, "loss": 1.7988, "step": 1721 }, { "epoch": 0.17, "learning_rate": 0.0018993795689386955, "loss": 1.6904, "step": 1722 }, { "epoch": 0.17, "learning_rate": 0.0018992399351599122, "loss": 1.8564, "step": 1723 }, { "epoch": 0.17, "learning_rate": 0.0018991002097013818, "loss": 1.8125, "step": 1724 }, { "epoch": 0.17, "learning_rate": 0.00189896039257735, "loss": 1.8516, "step": 1725 }, { "epoch": 0.17, "learning_rate": 0.0018988204838020712, "loss": 1.7324, "step": 1726 }, { "epoch": 0.17, "learning_rate": 0.0018986804833898093, "loss": 1.7939, "step": 1727 }, { "epoch": 0.17, "learning_rate": 0.0018985403913548379, "loss": 1.6885, "step": 1728 }, { "epoch": 0.17, "learning_rate": 0.0018984002077114398, "loss": 1.7334, "step": 1729 }, { "epoch": 0.17, "learning_rate": 0.001898259932473907, "loss": 1.917, "step": 1730 }, { "epoch": 0.17, "learning_rate": 0.0018981195656565407, "loss": 1.8818, "step": 1731 }, { "epoch": 0.17, "learning_rate": 0.0018979791072736521, "loss": 1.7627, "step": 1732 }, { "epoch": 0.17, "learning_rate": 0.001897838557339561, "loss": 1.792, "step": 1733 }, { "epoch": 0.17, "learning_rate": 0.0018976979158685965, "loss": 1.8965, "step": 1734 }, { "epoch": 0.17, "learning_rate": 0.0018975571828750981, "loss": 1.8965, "step": 1735 }, { "epoch": 0.17, "learning_rate": 0.0018974163583734135, "loss": 1.8652, "step": 1736 }, { "epoch": 0.17, "learning_rate": 0.0018972754423778996, "loss": 1.8125, "step": 1737 }, { "epoch": 0.17, "learning_rate": 0.0018971344349029242, "loss": 1.8604, "step": 1738 }, { "epoch": 0.17, "learning_rate": 0.0018969933359628626, "loss": 1.7891, "step": 1739 }, { "epoch": 0.17, "learning_rate": 0.0018968521455721003, "loss": 1.874, "step": 1740 }, { "epoch": 0.17, "learning_rate": 0.001896710863745032, "loss": 1.8525, "step": 1741 }, { "epoch": 0.17, "learning_rate": 0.0018965694904960618, "loss": 1.9424, "step": 1742 }, { "epoch": 0.17, "learning_rate": 0.0018964280258396032, "loss": 1.9395, "step": 1743 }, { "epoch": 0.17, "learning_rate": 0.0018962864697900786, "loss": 1.8037, "step": 1744 }, { "epoch": 0.17, "learning_rate": 0.0018961448223619203, "loss": 1.9092, "step": 1745 }, { "epoch": 0.17, "learning_rate": 0.0018960030835695693, "loss": 1.8682, "step": 1746 }, { "epoch": 0.17, "learning_rate": 0.0018958612534274765, "loss": 1.8936, "step": 1747 }, { "epoch": 0.17, "learning_rate": 0.0018957193319501014, "loss": 1.8965, "step": 1748 }, { "epoch": 0.17, "learning_rate": 0.0018955773191519137, "loss": 1.7617, "step": 1749 }, { "epoch": 0.17, "learning_rate": 0.001895435215047392, "loss": 1.8584, "step": 1750 }, { "epoch": 0.17, "learning_rate": 0.0018952930196510236, "loss": 1.7578, "step": 1751 }, { "epoch": 0.17, "learning_rate": 0.0018951507329773062, "loss": 1.7207, "step": 1752 }, { "epoch": 0.17, "learning_rate": 0.0018950083550407463, "loss": 1.6123, "step": 1753 }, { "epoch": 0.17, "learning_rate": 0.001894865885855859, "loss": 1.8457, "step": 1754 }, { "epoch": 0.17, "learning_rate": 0.0018947233254371706, "loss": 1.8828, "step": 1755 }, { "epoch": 0.17, "learning_rate": 0.0018945806737992142, "loss": 1.8545, "step": 1756 }, { "epoch": 0.17, "learning_rate": 0.0018944379309565344, "loss": 1.8398, "step": 1757 }, { "epoch": 0.17, "learning_rate": 0.0018942950969236835, "loss": 1.8154, "step": 1758 }, { "epoch": 0.17, "learning_rate": 0.0018941521717152246, "loss": 1.7568, "step": 1759 }, { "epoch": 0.17, "learning_rate": 0.0018940091553457286, "loss": 1.6885, "step": 1760 }, { "epoch": 0.17, "learning_rate": 0.0018938660478297767, "loss": 1.8682, "step": 1761 }, { "epoch": 0.17, "learning_rate": 0.001893722849181959, "loss": 1.8213, "step": 1762 }, { "epoch": 0.17, "learning_rate": 0.0018935795594168746, "loss": 1.8535, "step": 1763 }, { "epoch": 0.17, "learning_rate": 0.0018934361785491328, "loss": 1.8369, "step": 1764 }, { "epoch": 0.17, "learning_rate": 0.001893292706593351, "loss": 1.7852, "step": 1765 }, { "epoch": 0.17, "learning_rate": 0.0018931491435641577, "loss": 1.7949, "step": 1766 }, { "epoch": 0.17, "learning_rate": 0.0018930054894761882, "loss": 1.8936, "step": 1767 }, { "epoch": 0.17, "learning_rate": 0.001892861744344089, "loss": 1.8535, "step": 1768 }, { "epoch": 0.17, "learning_rate": 0.0018927179081825152, "loss": 1.8164, "step": 1769 }, { "epoch": 0.17, "learning_rate": 0.0018925739810061316, "loss": 1.7344, "step": 1770 }, { "epoch": 0.17, "learning_rate": 0.0018924299628296111, "loss": 1.6943, "step": 1771 }, { "epoch": 0.17, "learning_rate": 0.0018922858536676375, "loss": 1.8086, "step": 1772 }, { "epoch": 0.17, "learning_rate": 0.0018921416535349028, "loss": 1.8066, "step": 1773 }, { "epoch": 0.17, "learning_rate": 0.0018919973624461084, "loss": 1.7471, "step": 1774 }, { "epoch": 0.17, "learning_rate": 0.0018918529804159655, "loss": 1.9131, "step": 1775 }, { "epoch": 0.18, "learning_rate": 0.0018917085074591938, "loss": 1.8223, "step": 1776 }, { "epoch": 0.18, "learning_rate": 0.0018915639435905227, "loss": 1.7783, "step": 1777 }, { "epoch": 0.18, "learning_rate": 0.0018914192888246912, "loss": 1.7227, "step": 1778 }, { "epoch": 0.18, "learning_rate": 0.001891274543176447, "loss": 1.8281, "step": 1779 }, { "epoch": 0.18, "learning_rate": 0.0018911297066605472, "loss": 1.8262, "step": 1780 }, { "epoch": 0.18, "learning_rate": 0.0018909847792917588, "loss": 1.7842, "step": 1781 }, { "epoch": 0.18, "learning_rate": 0.0018908397610848567, "loss": 1.8447, "step": 1782 }, { "epoch": 0.18, "learning_rate": 0.0018906946520546263, "loss": 1.6875, "step": 1783 }, { "epoch": 0.18, "learning_rate": 0.0018905494522158618, "loss": 1.709, "step": 1784 }, { "epoch": 0.18, "learning_rate": 0.0018904041615833665, "loss": 1.7412, "step": 1785 }, { "epoch": 0.18, "learning_rate": 0.001890258780171953, "loss": 1.7539, "step": 1786 }, { "epoch": 0.18, "learning_rate": 0.001890113307996444, "loss": 1.8594, "step": 1787 }, { "epoch": 0.18, "learning_rate": 0.0018899677450716704, "loss": 1.7783, "step": 1788 }, { "epoch": 0.18, "learning_rate": 0.0018898220914124724, "loss": 1.8848, "step": 1789 }, { "epoch": 0.18, "learning_rate": 0.0018896763470337, "loss": 1.8311, "step": 1790 }, { "epoch": 0.18, "learning_rate": 0.0018895305119502122, "loss": 1.8096, "step": 1791 }, { "epoch": 0.18, "learning_rate": 0.001889384586176877, "loss": 1.7852, "step": 1792 }, { "epoch": 0.18, "learning_rate": 0.0018892385697285726, "loss": 1.6855, "step": 1793 }, { "epoch": 0.18, "learning_rate": 0.0018890924626201848, "loss": 1.8057, "step": 1794 }, { "epoch": 0.18, "learning_rate": 0.0018889462648666103, "loss": 1.7979, "step": 1795 }, { "epoch": 0.18, "learning_rate": 0.0018887999764827538, "loss": 1.8701, "step": 1796 }, { "epoch": 0.18, "learning_rate": 0.0018886535974835304, "loss": 1.7666, "step": 1797 }, { "epoch": 0.18, "learning_rate": 0.0018885071278838632, "loss": 1.7373, "step": 1798 }, { "epoch": 0.18, "learning_rate": 0.0018883605676986855, "loss": 1.9062, "step": 1799 }, { "epoch": 0.18, "learning_rate": 0.0018882139169429392, "loss": 1.8467, "step": 1800 }, { "epoch": 0.18, "learning_rate": 0.001888067175631576, "loss": 1.7607, "step": 1801 }, { "epoch": 0.18, "learning_rate": 0.0018879203437795564, "loss": 1.7695, "step": 1802 }, { "epoch": 0.18, "learning_rate": 0.0018877734214018502, "loss": 1.6719, "step": 1803 }, { "epoch": 0.18, "learning_rate": 0.0018876264085134367, "loss": 1.7754, "step": 1804 }, { "epoch": 0.18, "learning_rate": 0.001887479305129304, "loss": 1.7471, "step": 1805 }, { "epoch": 0.18, "learning_rate": 0.00188733211126445, "loss": 1.8271, "step": 1806 }, { "epoch": 0.18, "learning_rate": 0.001887184826933881, "loss": 1.8867, "step": 1807 }, { "epoch": 0.18, "learning_rate": 0.001887037452152613, "loss": 1.8623, "step": 1808 }, { "epoch": 0.18, "learning_rate": 0.001886889986935672, "loss": 1.8486, "step": 1809 }, { "epoch": 0.18, "learning_rate": 0.0018867424312980915, "loss": 1.791, "step": 1810 }, { "epoch": 0.18, "learning_rate": 0.0018865947852549157, "loss": 1.8408, "step": 1811 }, { "epoch": 0.18, "learning_rate": 0.0018864470488211974, "loss": 1.791, "step": 1812 }, { "epoch": 0.18, "learning_rate": 0.0018862992220119985, "loss": 1.7285, "step": 1813 }, { "epoch": 0.18, "learning_rate": 0.0018861513048423909, "loss": 1.8467, "step": 1814 }, { "epoch": 0.18, "learning_rate": 0.0018860032973274541, "loss": 1.9316, "step": 1815 }, { "epoch": 0.18, "learning_rate": 0.001885855199482279, "loss": 1.8643, "step": 1816 }, { "epoch": 0.18, "learning_rate": 0.0018857070113219635, "loss": 1.9482, "step": 1817 }, { "epoch": 0.18, "learning_rate": 0.0018855587328616163, "loss": 1.9131, "step": 1818 }, { "epoch": 0.18, "learning_rate": 0.0018854103641163548, "loss": 1.749, "step": 1819 }, { "epoch": 0.18, "learning_rate": 0.0018852619051013056, "loss": 1.9668, "step": 1820 }, { "epoch": 0.18, "learning_rate": 0.0018851133558316043, "loss": 1.7471, "step": 1821 }, { "epoch": 0.18, "learning_rate": 0.0018849647163223955, "loss": 1.6709, "step": 1822 }, { "epoch": 0.18, "learning_rate": 0.0018848159865888338, "loss": 1.8438, "step": 1823 }, { "epoch": 0.18, "learning_rate": 0.0018846671666460828, "loss": 1.8652, "step": 1824 }, { "epoch": 0.18, "learning_rate": 0.0018845182565093146, "loss": 1.8984, "step": 1825 }, { "epoch": 0.18, "learning_rate": 0.001884369256193711, "loss": 1.7998, "step": 1826 }, { "epoch": 0.18, "learning_rate": 0.0018842201657144635, "loss": 1.8105, "step": 1827 }, { "epoch": 0.18, "learning_rate": 0.0018840709850867718, "loss": 1.8242, "step": 1828 }, { "epoch": 0.18, "learning_rate": 0.0018839217143258448, "loss": 1.8457, "step": 1829 }, { "epoch": 0.18, "learning_rate": 0.0018837723534469017, "loss": 1.7666, "step": 1830 }, { "epoch": 0.18, "learning_rate": 0.00188362290246517, "loss": 1.8203, "step": 1831 }, { "epoch": 0.18, "learning_rate": 0.0018834733613958866, "loss": 1.8047, "step": 1832 }, { "epoch": 0.18, "learning_rate": 0.001883323730254298, "loss": 1.8789, "step": 1833 }, { "epoch": 0.18, "learning_rate": 0.0018831740090556584, "loss": 1.8164, "step": 1834 }, { "epoch": 0.18, "learning_rate": 0.0018830241978152332, "loss": 1.7725, "step": 1835 }, { "epoch": 0.18, "learning_rate": 0.0018828742965482953, "loss": 1.8125, "step": 1836 }, { "epoch": 0.18, "learning_rate": 0.0018827243052701285, "loss": 1.7061, "step": 1837 }, { "epoch": 0.18, "learning_rate": 0.0018825742239960241, "loss": 1.8535, "step": 1838 }, { "epoch": 0.18, "learning_rate": 0.0018824240527412833, "loss": 1.8398, "step": 1839 }, { "epoch": 0.18, "learning_rate": 0.0018822737915212167, "loss": 1.9658, "step": 1840 }, { "epoch": 0.18, "learning_rate": 0.0018821234403511432, "loss": 1.7285, "step": 1841 }, { "epoch": 0.18, "learning_rate": 0.001881972999246392, "loss": 1.8125, "step": 1842 }, { "epoch": 0.18, "learning_rate": 0.001881822468222301, "loss": 1.7041, "step": 1843 }, { "epoch": 0.18, "learning_rate": 0.0018816718472942171, "loss": 1.8242, "step": 1844 }, { "epoch": 0.18, "learning_rate": 0.0018815211364774966, "loss": 1.8193, "step": 1845 }, { "epoch": 0.18, "learning_rate": 0.0018813703357875042, "loss": 1.8125, "step": 1846 }, { "epoch": 0.18, "learning_rate": 0.001881219445239615, "loss": 1.8545, "step": 1847 }, { "epoch": 0.18, "learning_rate": 0.0018810684648492128, "loss": 1.7568, "step": 1848 }, { "epoch": 0.18, "learning_rate": 0.0018809173946316897, "loss": 1.7422, "step": 1849 }, { "epoch": 0.18, "learning_rate": 0.0018807662346024487, "loss": 1.7588, "step": 1850 }, { "epoch": 0.18, "learning_rate": 0.0018806149847769, "loss": 1.8076, "step": 1851 }, { "epoch": 0.18, "learning_rate": 0.0018804636451704643, "loss": 1.8301, "step": 1852 }, { "epoch": 0.18, "learning_rate": 0.0018803122157985712, "loss": 1.8213, "step": 1853 }, { "epoch": 0.18, "learning_rate": 0.0018801606966766591, "loss": 1.7725, "step": 1854 }, { "epoch": 0.18, "learning_rate": 0.0018800090878201758, "loss": 1.7832, "step": 1855 }, { "epoch": 0.18, "learning_rate": 0.0018798573892445784, "loss": 1.8301, "step": 1856 }, { "epoch": 0.18, "learning_rate": 0.0018797056009653321, "loss": 1.8086, "step": 1857 }, { "epoch": 0.18, "learning_rate": 0.0018795537229979132, "loss": 1.8311, "step": 1858 }, { "epoch": 0.18, "learning_rate": 0.0018794017553578055, "loss": 1.7275, "step": 1859 }, { "epoch": 0.18, "learning_rate": 0.0018792496980605026, "loss": 1.8623, "step": 1860 }, { "epoch": 0.18, "learning_rate": 0.0018790975511215069, "loss": 1.8203, "step": 1861 }, { "epoch": 0.18, "learning_rate": 0.0018789453145563303, "loss": 1.8262, "step": 1862 }, { "epoch": 0.18, "learning_rate": 0.001878792988380494, "loss": 1.7236, "step": 1863 }, { "epoch": 0.18, "learning_rate": 0.0018786405726095275, "loss": 1.8438, "step": 1864 }, { "epoch": 0.18, "learning_rate": 0.0018784880672589703, "loss": 1.8203, "step": 1865 }, { "epoch": 0.18, "learning_rate": 0.0018783354723443707, "loss": 1.6826, "step": 1866 }, { "epoch": 0.18, "learning_rate": 0.0018781827878812858, "loss": 1.8291, "step": 1867 }, { "epoch": 0.18, "learning_rate": 0.0018780300138852826, "loss": 1.8906, "step": 1868 }, { "epoch": 0.18, "learning_rate": 0.0018778771503719365, "loss": 1.8271, "step": 1869 }, { "epoch": 0.18, "learning_rate": 0.0018777241973568324, "loss": 1.8584, "step": 1870 }, { "epoch": 0.18, "learning_rate": 0.0018775711548555643, "loss": 1.6631, "step": 1871 }, { "epoch": 0.18, "learning_rate": 0.0018774180228837351, "loss": 1.8301, "step": 1872 }, { "epoch": 0.18, "learning_rate": 0.0018772648014569572, "loss": 1.9424, "step": 1873 }, { "epoch": 0.18, "learning_rate": 0.001877111490590852, "loss": 1.7617, "step": 1874 }, { "epoch": 0.18, "learning_rate": 0.0018769580903010494, "loss": 1.833, "step": 1875 }, { "epoch": 0.18, "learning_rate": 0.0018768046006031895, "loss": 1.7529, "step": 1876 }, { "epoch": 0.19, "learning_rate": 0.0018766510215129206, "loss": 1.8867, "step": 1877 }, { "epoch": 0.19, "learning_rate": 0.0018764973530459006, "loss": 1.7441, "step": 1878 }, { "epoch": 0.19, "learning_rate": 0.0018763435952177964, "loss": 1.8652, "step": 1879 }, { "epoch": 0.19, "learning_rate": 0.001876189748044284, "loss": 1.7754, "step": 1880 }, { "epoch": 0.19, "learning_rate": 0.0018760358115410486, "loss": 1.7344, "step": 1881 }, { "epoch": 0.19, "learning_rate": 0.001875881785723784, "loss": 1.918, "step": 1882 }, { "epoch": 0.19, "learning_rate": 0.001875727670608194, "loss": 1.9102, "step": 1883 }, { "epoch": 0.19, "learning_rate": 0.0018755734662099907, "loss": 1.7471, "step": 1884 }, { "epoch": 0.19, "learning_rate": 0.0018754191725448956, "loss": 1.8047, "step": 1885 }, { "epoch": 0.19, "learning_rate": 0.0018752647896286398, "loss": 1.7734, "step": 1886 }, { "epoch": 0.19, "learning_rate": 0.0018751103174769628, "loss": 1.6865, "step": 1887 }, { "epoch": 0.19, "learning_rate": 0.0018749557561056128, "loss": 1.8916, "step": 1888 }, { "epoch": 0.19, "learning_rate": 0.0018748011055303487, "loss": 1.8516, "step": 1889 }, { "epoch": 0.19, "learning_rate": 0.0018746463657669365, "loss": 1.6777, "step": 1890 }, { "epoch": 0.19, "learning_rate": 0.0018744915368311531, "loss": 1.8193, "step": 1891 }, { "epoch": 0.19, "learning_rate": 0.0018743366187387836, "loss": 1.7959, "step": 1892 }, { "epoch": 0.19, "learning_rate": 0.001874181611505622, "loss": 1.7949, "step": 1893 }, { "epoch": 0.19, "learning_rate": 0.0018740265151474715, "loss": 1.8213, "step": 1894 }, { "epoch": 0.19, "learning_rate": 0.0018738713296801452, "loss": 1.6875, "step": 1895 }, { "epoch": 0.19, "learning_rate": 0.001873716055119464, "loss": 1.958, "step": 1896 }, { "epoch": 0.19, "learning_rate": 0.0018735606914812592, "loss": 1.7764, "step": 1897 }, { "epoch": 0.19, "learning_rate": 0.0018734052387813698, "loss": 1.7803, "step": 1898 }, { "epoch": 0.19, "learning_rate": 0.0018732496970356451, "loss": 1.7734, "step": 1899 }, { "epoch": 0.19, "learning_rate": 0.0018730940662599427, "loss": 1.832, "step": 1900 }, { "epoch": 0.19, "learning_rate": 0.0018729383464701297, "loss": 1.8623, "step": 1901 }, { "epoch": 0.19, "learning_rate": 0.001872782537682082, "loss": 1.7402, "step": 1902 }, { "epoch": 0.19, "learning_rate": 0.0018726266399116847, "loss": 1.8623, "step": 1903 }, { "epoch": 0.19, "learning_rate": 0.0018724706531748321, "loss": 1.915, "step": 1904 }, { "epoch": 0.19, "learning_rate": 0.0018723145774874274, "loss": 1.7422, "step": 1905 }, { "epoch": 0.19, "learning_rate": 0.0018721584128653826, "loss": 1.8428, "step": 1906 }, { "epoch": 0.19, "learning_rate": 0.0018720021593246195, "loss": 1.8301, "step": 1907 }, { "epoch": 0.19, "learning_rate": 0.0018718458168810682, "loss": 1.7852, "step": 1908 }, { "epoch": 0.19, "learning_rate": 0.0018716893855506682, "loss": 1.7021, "step": 1909 }, { "epoch": 0.19, "learning_rate": 0.0018715328653493685, "loss": 1.8682, "step": 1910 }, { "epoch": 0.19, "learning_rate": 0.0018713762562931264, "loss": 1.8164, "step": 1911 }, { "epoch": 0.19, "learning_rate": 0.0018712195583979088, "loss": 1.8008, "step": 1912 }, { "epoch": 0.19, "learning_rate": 0.001871062771679691, "loss": 1.7949, "step": 1913 }, { "epoch": 0.19, "learning_rate": 0.0018709058961544583, "loss": 1.7744, "step": 1914 }, { "epoch": 0.19, "learning_rate": 0.001870748931838204, "loss": 1.8545, "step": 1915 }, { "epoch": 0.19, "learning_rate": 0.0018705918787469316, "loss": 1.8193, "step": 1916 }, { "epoch": 0.19, "learning_rate": 0.0018704347368966525, "loss": 1.8955, "step": 1917 }, { "epoch": 0.19, "learning_rate": 0.0018702775063033884, "loss": 1.7012, "step": 1918 }, { "epoch": 0.19, "learning_rate": 0.0018701201869831684, "loss": 1.7783, "step": 1919 }, { "epoch": 0.19, "learning_rate": 0.0018699627789520324, "loss": 1.8271, "step": 1920 }, { "epoch": 0.19, "learning_rate": 0.0018698052822260281, "loss": 1.8809, "step": 1921 }, { "epoch": 0.19, "learning_rate": 0.0018696476968212128, "loss": 1.8984, "step": 1922 }, { "epoch": 0.19, "learning_rate": 0.001869490022753653, "loss": 1.8672, "step": 1923 }, { "epoch": 0.19, "learning_rate": 0.0018693322600394236, "loss": 1.8232, "step": 1924 }, { "epoch": 0.19, "learning_rate": 0.001869174408694609, "loss": 1.7305, "step": 1925 }, { "epoch": 0.19, "learning_rate": 0.0018690164687353026, "loss": 1.5947, "step": 1926 }, { "epoch": 0.19, "learning_rate": 0.0018688584401776068, "loss": 1.8125, "step": 1927 }, { "epoch": 0.19, "learning_rate": 0.0018687003230376327, "loss": 1.7607, "step": 1928 }, { "epoch": 0.19, "learning_rate": 0.0018685421173315012, "loss": 1.8184, "step": 1929 }, { "epoch": 0.19, "learning_rate": 0.0018683838230753413, "loss": 1.7656, "step": 1930 }, { "epoch": 0.19, "learning_rate": 0.001868225440285292, "loss": 1.7402, "step": 1931 }, { "epoch": 0.19, "learning_rate": 0.0018680669689775003, "loss": 1.7754, "step": 1932 }, { "epoch": 0.19, "learning_rate": 0.0018679084091681234, "loss": 1.8984, "step": 1933 }, { "epoch": 0.19, "learning_rate": 0.001867749760873326, "loss": 1.585, "step": 1934 }, { "epoch": 0.19, "learning_rate": 0.0018675910241092834, "loss": 1.7578, "step": 1935 }, { "epoch": 0.19, "learning_rate": 0.0018674321988921788, "loss": 1.7695, "step": 1936 }, { "epoch": 0.19, "learning_rate": 0.001867273285238205, "loss": 1.8467, "step": 1937 }, { "epoch": 0.19, "learning_rate": 0.0018671142831635638, "loss": 1.7002, "step": 1938 }, { "epoch": 0.19, "learning_rate": 0.0018669551926844656, "loss": 1.8047, "step": 1939 }, { "epoch": 0.19, "learning_rate": 0.00186679601381713, "loss": 1.8311, "step": 1940 }, { "epoch": 0.19, "learning_rate": 0.001866636746577786, "loss": 1.7012, "step": 1941 }, { "epoch": 0.19, "learning_rate": 0.001866477390982671, "loss": 1.666, "step": 1942 }, { "epoch": 0.19, "learning_rate": 0.001866317947048032, "loss": 1.8047, "step": 1943 }, { "epoch": 0.19, "learning_rate": 0.0018661584147901245, "loss": 1.8037, "step": 1944 }, { "epoch": 0.19, "learning_rate": 0.001865998794225213, "loss": 1.7402, "step": 1945 }, { "epoch": 0.19, "learning_rate": 0.0018658390853695717, "loss": 1.9033, "step": 1946 }, { "epoch": 0.19, "learning_rate": 0.0018656792882394831, "loss": 1.793, "step": 1947 }, { "epoch": 0.19, "learning_rate": 0.0018655194028512389, "loss": 1.8516, "step": 1948 }, { "epoch": 0.19, "learning_rate": 0.0018653594292211393, "loss": 1.8975, "step": 1949 }, { "epoch": 0.19, "learning_rate": 0.001865199367365495, "loss": 1.8535, "step": 1950 }, { "epoch": 0.19, "learning_rate": 0.001865039217300624, "loss": 1.8291, "step": 1951 }, { "epoch": 0.19, "learning_rate": 0.0018648789790428543, "loss": 1.71, "step": 1952 }, { "epoch": 0.19, "learning_rate": 0.0018647186526085224, "loss": 1.8418, "step": 1953 }, { "epoch": 0.19, "learning_rate": 0.001864558238013974, "loss": 1.8418, "step": 1954 }, { "epoch": 0.19, "learning_rate": 0.0018643977352755642, "loss": 1.7207, "step": 1955 }, { "epoch": 0.19, "learning_rate": 0.001864237144409656, "loss": 1.8486, "step": 1956 }, { "epoch": 0.19, "learning_rate": 0.0018640764654326224, "loss": 1.7656, "step": 1957 }, { "epoch": 0.19, "learning_rate": 0.001863915698360845, "loss": 1.7842, "step": 1958 }, { "epoch": 0.19, "learning_rate": 0.0018637548432107144, "loss": 1.9248, "step": 1959 }, { "epoch": 0.19, "learning_rate": 0.0018635938999986299, "loss": 1.877, "step": 1960 }, { "epoch": 0.19, "learning_rate": 0.0018634328687410005, "loss": 1.7764, "step": 1961 }, { "epoch": 0.19, "learning_rate": 0.0018632717494542434, "loss": 1.7588, "step": 1962 }, { "epoch": 0.19, "learning_rate": 0.0018631105421547853, "loss": 1.8486, "step": 1963 }, { "epoch": 0.19, "learning_rate": 0.0018629492468590616, "loss": 1.8271, "step": 1964 }, { "epoch": 0.19, "learning_rate": 0.0018627878635835166, "loss": 1.7305, "step": 1965 }, { "epoch": 0.19, "learning_rate": 0.0018626263923446043, "loss": 1.6807, "step": 1966 }, { "epoch": 0.19, "learning_rate": 0.0018624648331587866, "loss": 1.7832, "step": 1967 }, { "epoch": 0.19, "learning_rate": 0.0018623031860425347, "loss": 1.7129, "step": 1968 }, { "epoch": 0.19, "learning_rate": 0.0018621414510123296, "loss": 1.7598, "step": 1969 }, { "epoch": 0.19, "learning_rate": 0.0018619796280846597, "loss": 1.8203, "step": 1970 }, { "epoch": 0.19, "learning_rate": 0.0018618177172760242, "loss": 1.8135, "step": 1971 }, { "epoch": 0.19, "learning_rate": 0.0018616557186029298, "loss": 1.8174, "step": 1972 }, { "epoch": 0.19, "learning_rate": 0.0018614936320818926, "loss": 1.8135, "step": 1973 }, { "epoch": 0.19, "learning_rate": 0.0018613314577294376, "loss": 1.8447, "step": 1974 }, { "epoch": 0.19, "learning_rate": 0.0018611691955620993, "loss": 1.7305, "step": 1975 }, { "epoch": 0.19, "learning_rate": 0.0018610068455964207, "loss": 1.8125, "step": 1976 }, { "epoch": 0.19, "learning_rate": 0.0018608444078489533, "loss": 1.8701, "step": 1977 }, { "epoch": 0.19, "learning_rate": 0.0018606818823362584, "loss": 1.6777, "step": 1978 }, { "epoch": 0.2, "learning_rate": 0.001860519269074906, "loss": 1.7979, "step": 1979 }, { "epoch": 0.2, "learning_rate": 0.0018603565680814747, "loss": 1.7773, "step": 1980 }, { "epoch": 0.2, "learning_rate": 0.0018601937793725524, "loss": 1.7959, "step": 1981 }, { "epoch": 0.2, "learning_rate": 0.0018600309029647355, "loss": 1.9141, "step": 1982 }, { "epoch": 0.2, "learning_rate": 0.00185986793887463, "loss": 1.7861, "step": 1983 }, { "epoch": 0.2, "learning_rate": 0.0018597048871188506, "loss": 1.8477, "step": 1984 }, { "epoch": 0.2, "learning_rate": 0.0018595417477140206, "loss": 1.8145, "step": 1985 }, { "epoch": 0.2, "learning_rate": 0.0018593785206767724, "loss": 1.8623, "step": 1986 }, { "epoch": 0.2, "learning_rate": 0.0018592152060237475, "loss": 1.8301, "step": 1987 }, { "epoch": 0.2, "learning_rate": 0.0018590518037715964, "loss": 1.835, "step": 1988 }, { "epoch": 0.2, "learning_rate": 0.001858888313936978, "loss": 1.7881, "step": 1989 }, { "epoch": 0.2, "learning_rate": 0.0018587247365365607, "loss": 1.7383, "step": 1990 }, { "epoch": 0.2, "learning_rate": 0.0018585610715870216, "loss": 1.9375, "step": 1991 }, { "epoch": 0.2, "learning_rate": 0.0018583973191050468, "loss": 1.7578, "step": 1992 }, { "epoch": 0.2, "learning_rate": 0.0018582334791073312, "loss": 1.7275, "step": 1993 }, { "epoch": 0.2, "learning_rate": 0.0018580695516105788, "loss": 1.7734, "step": 1994 }, { "epoch": 0.2, "learning_rate": 0.0018579055366315022, "loss": 1.8965, "step": 1995 }, { "epoch": 0.2, "learning_rate": 0.0018577414341868234, "loss": 1.8486, "step": 1996 }, { "epoch": 0.2, "learning_rate": 0.0018575772442932733, "loss": 1.8105, "step": 1997 }, { "epoch": 0.2, "learning_rate": 0.0018574129669675905, "loss": 1.7578, "step": 1998 }, { "epoch": 0.2, "learning_rate": 0.0018572486022265244, "loss": 1.8174, "step": 1999 }, { "epoch": 0.2, "learning_rate": 0.0018570841500868318, "loss": 1.8623, "step": 2000 }, { "epoch": 0.2, "learning_rate": 0.0018569196105652796, "loss": 1.8584, "step": 2001 }, { "epoch": 0.2, "learning_rate": 0.0018567549836786426, "loss": 1.8262, "step": 2002 }, { "epoch": 0.2, "learning_rate": 0.0018565902694437052, "loss": 1.7529, "step": 2003 }, { "epoch": 0.2, "learning_rate": 0.0018564254678772598, "loss": 1.707, "step": 2004 }, { "epoch": 0.2, "learning_rate": 0.001856260578996109, "loss": 1.832, "step": 2005 }, { "epoch": 0.2, "learning_rate": 0.0018560956028170637, "loss": 1.873, "step": 2006 }, { "epoch": 0.2, "learning_rate": 0.001855930539356943, "loss": 1.7217, "step": 2007 }, { "epoch": 0.2, "learning_rate": 0.0018557653886325759, "loss": 1.8213, "step": 2008 }, { "epoch": 0.2, "learning_rate": 0.0018556001506607998, "loss": 1.707, "step": 2009 }, { "epoch": 0.2, "learning_rate": 0.0018554348254584613, "loss": 1.7715, "step": 2010 }, { "epoch": 0.2, "learning_rate": 0.0018552694130424157, "loss": 1.6943, "step": 2011 }, { "epoch": 0.2, "learning_rate": 0.0018551039134295274, "loss": 1.8232, "step": 2012 }, { "epoch": 0.2, "learning_rate": 0.0018549383266366687, "loss": 1.6904, "step": 2013 }, { "epoch": 0.2, "learning_rate": 0.0018547726526807224, "loss": 1.6748, "step": 2014 }, { "epoch": 0.2, "learning_rate": 0.001854606891578579, "loss": 1.7471, "step": 2015 }, { "epoch": 0.2, "learning_rate": 0.0018544410433471385, "loss": 1.7617, "step": 2016 }, { "epoch": 0.2, "learning_rate": 0.0018542751080033093, "loss": 1.8604, "step": 2017 }, { "epoch": 0.2, "learning_rate": 0.001854109085564009, "loss": 1.75, "step": 2018 }, { "epoch": 0.2, "learning_rate": 0.001853942976046164, "loss": 1.8145, "step": 2019 }, { "epoch": 0.2, "learning_rate": 0.0018537767794667096, "loss": 1.8389, "step": 2020 }, { "epoch": 0.2, "learning_rate": 0.00185361049584259, "loss": 1.8652, "step": 2021 }, { "epoch": 0.2, "learning_rate": 0.0018534441251907578, "loss": 1.6992, "step": 2022 }, { "epoch": 0.2, "learning_rate": 0.0018532776675281757, "loss": 1.8926, "step": 2023 }, { "epoch": 0.2, "learning_rate": 0.0018531111228718138, "loss": 1.7861, "step": 2024 }, { "epoch": 0.2, "learning_rate": 0.001852944491238652, "loss": 1.8428, "step": 2025 }, { "epoch": 0.2, "learning_rate": 0.0018527777726456789, "loss": 1.666, "step": 2026 }, { "epoch": 0.2, "learning_rate": 0.001852610967109892, "loss": 1.666, "step": 2027 }, { "epoch": 0.2, "learning_rate": 0.0018524440746482968, "loss": 1.7256, "step": 2028 }, { "epoch": 0.2, "learning_rate": 0.001852277095277909, "loss": 1.7393, "step": 2029 }, { "epoch": 0.2, "learning_rate": 0.0018521100290157528, "loss": 1.8105, "step": 2030 }, { "epoch": 0.2, "learning_rate": 0.0018519428758788608, "loss": 1.7881, "step": 2031 }, { "epoch": 0.2, "learning_rate": 0.001851775635884274, "loss": 1.7305, "step": 2032 }, { "epoch": 0.2, "learning_rate": 0.0018516083090490438, "loss": 1.7793, "step": 2033 }, { "epoch": 0.2, "learning_rate": 0.0018514408953902294, "loss": 1.7822, "step": 2034 }, { "epoch": 0.2, "learning_rate": 0.001851273394924899, "loss": 1.8975, "step": 2035 }, { "epoch": 0.2, "learning_rate": 0.0018511058076701296, "loss": 1.7549, "step": 2036 }, { "epoch": 0.2, "learning_rate": 0.0018509381336430067, "loss": 1.7109, "step": 2037 }, { "epoch": 0.2, "learning_rate": 0.0018507703728606257, "loss": 1.9053, "step": 2038 }, { "epoch": 0.2, "learning_rate": 0.0018506025253400905, "loss": 1.8711, "step": 2039 }, { "epoch": 0.2, "learning_rate": 0.0018504345910985128, "loss": 1.8262, "step": 2040 }, { "epoch": 0.2, "learning_rate": 0.0018502665701530141, "loss": 1.7119, "step": 2041 }, { "epoch": 0.2, "learning_rate": 0.001850098462520725, "loss": 1.7637, "step": 2042 }, { "epoch": 0.2, "learning_rate": 0.0018499302682187843, "loss": 1.8096, "step": 2043 }, { "epoch": 0.2, "learning_rate": 0.0018497619872643394, "loss": 1.8877, "step": 2044 }, { "epoch": 0.2, "learning_rate": 0.0018495936196745472, "loss": 1.7803, "step": 2045 }, { "epoch": 0.2, "learning_rate": 0.0018494251654665736, "loss": 1.7334, "step": 2046 }, { "epoch": 0.2, "learning_rate": 0.0018492566246575927, "loss": 1.8037, "step": 2047 }, { "epoch": 0.2, "learning_rate": 0.0018490879972647872, "loss": 1.8203, "step": 2048 }, { "epoch": 0.2, "learning_rate": 0.00184891928330535, "loss": 1.8838, "step": 2049 }, { "epoch": 0.2, "learning_rate": 0.0018487504827964804, "loss": 1.8691, "step": 2050 }, { "epoch": 0.2, "learning_rate": 0.0018485815957553896, "loss": 1.8115, "step": 2051 }, { "epoch": 0.2, "learning_rate": 0.0018484126221992956, "loss": 1.7812, "step": 2052 }, { "epoch": 0.2, "learning_rate": 0.0018482435621454255, "loss": 1.6992, "step": 2053 }, { "epoch": 0.2, "learning_rate": 0.001848074415611015, "loss": 1.7607, "step": 2054 }, { "epoch": 0.2, "learning_rate": 0.0018479051826133098, "loss": 1.79, "step": 2055 }, { "epoch": 0.2, "learning_rate": 0.0018477358631695628, "loss": 1.7891, "step": 2056 }, { "epoch": 0.2, "learning_rate": 0.0018475664572970372, "loss": 1.791, "step": 2057 }, { "epoch": 0.2, "learning_rate": 0.001847396965013004, "loss": 1.8564, "step": 2058 }, { "epoch": 0.2, "learning_rate": 0.001847227386334744, "loss": 1.6729, "step": 2059 }, { "epoch": 0.2, "learning_rate": 0.001847057721279545, "loss": 1.8027, "step": 2060 }, { "epoch": 0.2, "learning_rate": 0.0018468879698647059, "loss": 1.7373, "step": 2061 }, { "epoch": 0.2, "learning_rate": 0.0018467181321075326, "loss": 1.8701, "step": 2062 }, { "epoch": 0.2, "learning_rate": 0.0018465482080253406, "loss": 1.6865, "step": 2063 }, { "epoch": 0.2, "learning_rate": 0.0018463781976354543, "loss": 1.8398, "step": 2064 }, { "epoch": 0.2, "learning_rate": 0.0018462081009552064, "loss": 1.8438, "step": 2065 }, { "epoch": 0.2, "learning_rate": 0.001846037918001939, "loss": 1.7803, "step": 2066 }, { "epoch": 0.2, "learning_rate": 0.0018458676487930026, "loss": 1.7031, "step": 2067 }, { "epoch": 0.2, "learning_rate": 0.001845697293345756, "loss": 1.748, "step": 2068 }, { "epoch": 0.2, "learning_rate": 0.0018455268516775684, "loss": 1.749, "step": 2069 }, { "epoch": 0.2, "learning_rate": 0.0018453563238058162, "loss": 1.7998, "step": 2070 }, { "epoch": 0.2, "learning_rate": 0.0018451857097478849, "loss": 1.7256, "step": 2071 }, { "epoch": 0.2, "learning_rate": 0.001845015009521169, "loss": 1.7012, "step": 2072 }, { "epoch": 0.2, "learning_rate": 0.0018448442231430724, "loss": 1.6719, "step": 2073 }, { "epoch": 0.2, "learning_rate": 0.0018446733506310067, "loss": 1.8418, "step": 2074 }, { "epoch": 0.2, "learning_rate": 0.001844502392002393, "loss": 1.7891, "step": 2075 }, { "epoch": 0.2, "learning_rate": 0.001844331347274661, "loss": 1.7822, "step": 2076 }, { "epoch": 0.2, "learning_rate": 0.0018441602164652488, "loss": 1.7666, "step": 2077 }, { "epoch": 0.2, "learning_rate": 0.0018439889995916041, "loss": 1.75, "step": 2078 }, { "epoch": 0.2, "learning_rate": 0.0018438176966711825, "loss": 1.7861, "step": 2079 }, { "epoch": 0.21, "learning_rate": 0.001843646307721449, "loss": 1.8789, "step": 2080 }, { "epoch": 0.21, "learning_rate": 0.0018434748327598772, "loss": 1.7217, "step": 2081 }, { "epoch": 0.21, "learning_rate": 0.0018433032718039488, "loss": 1.7764, "step": 2082 }, { "epoch": 0.21, "learning_rate": 0.0018431316248711553, "loss": 1.7725, "step": 2083 }, { "epoch": 0.21, "learning_rate": 0.0018429598919789969, "loss": 1.8359, "step": 2084 }, { "epoch": 0.21, "learning_rate": 0.0018427880731449815, "loss": 1.8086, "step": 2085 }, { "epoch": 0.21, "learning_rate": 0.001842616168386627, "loss": 1.8623, "step": 2086 }, { "epoch": 0.21, "learning_rate": 0.0018424441777214588, "loss": 1.7217, "step": 2087 }, { "epoch": 0.21, "learning_rate": 0.001842272101167013, "loss": 1.75, "step": 2088 }, { "epoch": 0.21, "learning_rate": 0.0018420999387408318, "loss": 1.8525, "step": 2089 }, { "epoch": 0.21, "learning_rate": 0.0018419276904604687, "loss": 1.8535, "step": 2090 }, { "epoch": 0.21, "learning_rate": 0.0018417553563434843, "loss": 1.8701, "step": 2091 }, { "epoch": 0.21, "learning_rate": 0.0018415829364074486, "loss": 1.8916, "step": 2092 }, { "epoch": 0.21, "learning_rate": 0.0018414104306699404, "loss": 1.7451, "step": 2093 }, { "epoch": 0.21, "learning_rate": 0.0018412378391485468, "loss": 1.7256, "step": 2094 }, { "epoch": 0.21, "learning_rate": 0.0018410651618608641, "loss": 1.7744, "step": 2095 }, { "epoch": 0.21, "learning_rate": 0.0018408923988244973, "loss": 1.7861, "step": 2096 }, { "epoch": 0.21, "learning_rate": 0.0018407195500570596, "loss": 1.7686, "step": 2097 }, { "epoch": 0.21, "learning_rate": 0.0018405466155761738, "loss": 1.8584, "step": 2098 }, { "epoch": 0.21, "learning_rate": 0.0018403735953994708, "loss": 1.8789, "step": 2099 }, { "epoch": 0.21, "learning_rate": 0.0018402004895445906, "loss": 1.7109, "step": 2100 }, { "epoch": 0.21, "learning_rate": 0.0018400272980291814, "loss": 1.6982, "step": 2101 }, { "epoch": 0.21, "learning_rate": 0.0018398540208709009, "loss": 1.749, "step": 2102 }, { "epoch": 0.21, "learning_rate": 0.0018396806580874149, "loss": 1.6621, "step": 2103 }, { "epoch": 0.21, "learning_rate": 0.0018395072096963983, "loss": 1.7588, "step": 2104 }, { "epoch": 0.21, "learning_rate": 0.0018393336757155346, "loss": 1.7139, "step": 2105 }, { "epoch": 0.21, "learning_rate": 0.001839160056162516, "loss": 1.9102, "step": 2106 }, { "epoch": 0.21, "learning_rate": 0.0018389863510550435, "loss": 1.8301, "step": 2107 }, { "epoch": 0.21, "learning_rate": 0.0018388125604108265, "loss": 1.8408, "step": 2108 }, { "epoch": 0.21, "learning_rate": 0.0018386386842475838, "loss": 1.7461, "step": 2109 }, { "epoch": 0.21, "learning_rate": 0.001838464722583042, "loss": 1.7119, "step": 2110 }, { "epoch": 0.21, "learning_rate": 0.0018382906754349376, "loss": 1.6855, "step": 2111 }, { "epoch": 0.21, "learning_rate": 0.0018381165428210144, "loss": 1.8516, "step": 2112 }, { "epoch": 0.21, "learning_rate": 0.001837942324759026, "loss": 1.8789, "step": 2113 }, { "epoch": 0.21, "learning_rate": 0.0018377680212667347, "loss": 1.7168, "step": 2114 }, { "epoch": 0.21, "learning_rate": 0.0018375936323619104, "loss": 1.7246, "step": 2115 }, { "epoch": 0.21, "learning_rate": 0.0018374191580623333, "loss": 1.7822, "step": 2116 }, { "epoch": 0.21, "learning_rate": 0.0018372445983857908, "loss": 1.7959, "step": 2117 }, { "epoch": 0.21, "learning_rate": 0.0018370699533500803, "loss": 1.752, "step": 2118 }, { "epoch": 0.21, "learning_rate": 0.0018368952229730067, "loss": 1.9307, "step": 2119 }, { "epoch": 0.21, "learning_rate": 0.0018367204072723847, "loss": 1.8223, "step": 2120 }, { "epoch": 0.21, "learning_rate": 0.0018365455062660369, "loss": 1.8369, "step": 2121 }, { "epoch": 0.21, "learning_rate": 0.0018363705199717952, "loss": 1.7734, "step": 2122 }, { "epoch": 0.21, "learning_rate": 0.0018361954484074992, "loss": 1.8438, "step": 2123 }, { "epoch": 0.21, "learning_rate": 0.0018360202915909988, "loss": 1.7412, "step": 2124 }, { "epoch": 0.21, "learning_rate": 0.001835845049540151, "loss": 1.8115, "step": 2125 }, { "epoch": 0.21, "learning_rate": 0.0018356697222728223, "loss": 1.7832, "step": 2126 }, { "epoch": 0.21, "learning_rate": 0.001835494309806888, "loss": 1.8291, "step": 2127 }, { "epoch": 0.21, "learning_rate": 0.0018353188121602317, "loss": 1.7236, "step": 2128 }, { "epoch": 0.21, "learning_rate": 0.0018351432293507459, "loss": 1.8389, "step": 2129 }, { "epoch": 0.21, "learning_rate": 0.0018349675613963314, "loss": 1.7002, "step": 2130 }, { "epoch": 0.21, "learning_rate": 0.0018347918083148982, "loss": 1.7754, "step": 2131 }, { "epoch": 0.21, "learning_rate": 0.0018346159701243647, "loss": 1.7305, "step": 2132 }, { "epoch": 0.21, "learning_rate": 0.0018344400468426581, "loss": 1.8643, "step": 2133 }, { "epoch": 0.21, "learning_rate": 0.0018342640384877145, "loss": 1.7539, "step": 2134 }, { "epoch": 0.21, "learning_rate": 0.001834087945077478, "loss": 1.8076, "step": 2135 }, { "epoch": 0.21, "learning_rate": 0.0018339117666299014, "loss": 1.8311, "step": 2136 }, { "epoch": 0.21, "learning_rate": 0.0018337355031629473, "loss": 1.8086, "step": 2137 }, { "epoch": 0.21, "learning_rate": 0.001833559154694586, "loss": 1.7324, "step": 2138 }, { "epoch": 0.21, "learning_rate": 0.0018333827212427965, "loss": 1.5713, "step": 2139 }, { "epoch": 0.21, "learning_rate": 0.0018332062028255667, "loss": 1.7539, "step": 2140 }, { "epoch": 0.21, "learning_rate": 0.0018330295994608927, "loss": 1.7949, "step": 2141 }, { "epoch": 0.21, "learning_rate": 0.0018328529111667807, "loss": 1.7881, "step": 2142 }, { "epoch": 0.21, "learning_rate": 0.001832676137961243, "loss": 1.8174, "step": 2143 }, { "epoch": 0.21, "learning_rate": 0.0018324992798623035, "loss": 1.7734, "step": 2144 }, { "epoch": 0.21, "learning_rate": 0.0018323223368879927, "loss": 1.7676, "step": 2145 }, { "epoch": 0.21, "learning_rate": 0.0018321453090563503, "loss": 1.8428, "step": 2146 }, { "epoch": 0.21, "learning_rate": 0.0018319681963854246, "loss": 1.6191, "step": 2147 }, { "epoch": 0.21, "learning_rate": 0.0018317909988932732, "loss": 1.7529, "step": 2148 }, { "epoch": 0.21, "learning_rate": 0.0018316137165979615, "loss": 1.832, "step": 2149 }, { "epoch": 0.21, "learning_rate": 0.0018314363495175639, "loss": 1.7686, "step": 2150 }, { "epoch": 0.21, "learning_rate": 0.0018312588976701631, "loss": 1.8496, "step": 2151 }, { "epoch": 0.21, "learning_rate": 0.0018310813610738514, "loss": 1.8076, "step": 2152 }, { "epoch": 0.21, "learning_rate": 0.0018309037397467288, "loss": 1.7178, "step": 2153 }, { "epoch": 0.21, "learning_rate": 0.0018307260337069042, "loss": 1.8184, "step": 2154 }, { "epoch": 0.21, "learning_rate": 0.001830548242972495, "loss": 1.7285, "step": 2155 }, { "epoch": 0.21, "learning_rate": 0.0018303703675616276, "loss": 1.7451, "step": 2156 }, { "epoch": 0.21, "learning_rate": 0.001830192407492437, "loss": 1.8193, "step": 2157 }, { "epoch": 0.21, "learning_rate": 0.0018300143627830664, "loss": 1.7295, "step": 2158 }, { "epoch": 0.21, "learning_rate": 0.001829836233451668, "loss": 1.7676, "step": 2159 }, { "epoch": 0.21, "learning_rate": 0.0018296580195164022, "loss": 1.7451, "step": 2160 }, { "epoch": 0.21, "learning_rate": 0.001829479720995439, "loss": 1.8672, "step": 2161 }, { "epoch": 0.21, "learning_rate": 0.0018293013379069563, "loss": 1.7812, "step": 2162 }, { "epoch": 0.21, "learning_rate": 0.0018291228702691403, "loss": 1.7568, "step": 2163 }, { "epoch": 0.21, "learning_rate": 0.001828944318100186, "loss": 1.8438, "step": 2164 }, { "epoch": 0.21, "learning_rate": 0.001828765681418298, "loss": 1.9697, "step": 2165 }, { "epoch": 0.21, "learning_rate": 0.0018285869602416883, "loss": 1.6553, "step": 2166 }, { "epoch": 0.21, "learning_rate": 0.001828408154588578, "loss": 1.8018, "step": 2167 }, { "epoch": 0.21, "learning_rate": 0.0018282292644771967, "loss": 1.8213, "step": 2168 }, { "epoch": 0.21, "learning_rate": 0.001828050289925783, "loss": 1.8994, "step": 2169 }, { "epoch": 0.21, "learning_rate": 0.0018278712309525832, "loss": 1.708, "step": 2170 }, { "epoch": 0.21, "learning_rate": 0.0018276920875758538, "loss": 1.7188, "step": 2171 }, { "epoch": 0.21, "learning_rate": 0.001827512859813858, "loss": 1.752, "step": 2172 }, { "epoch": 0.21, "learning_rate": 0.001827333547684869, "loss": 1.7969, "step": 2173 }, { "epoch": 0.21, "learning_rate": 0.001827154151207168, "loss": 1.7725, "step": 2174 }, { "epoch": 0.21, "learning_rate": 0.0018269746703990447, "loss": 1.7441, "step": 2175 }, { "epoch": 0.21, "learning_rate": 0.0018267951052787978, "loss": 1.8594, "step": 2176 }, { "epoch": 0.21, "learning_rate": 0.0018266154558647344, "loss": 1.834, "step": 2177 }, { "epoch": 0.21, "learning_rate": 0.0018264357221751703, "loss": 1.6973, "step": 2178 }, { "epoch": 0.21, "learning_rate": 0.0018262559042284297, "loss": 1.8037, "step": 2179 }, { "epoch": 0.21, "learning_rate": 0.0018260760020428455, "loss": 1.7412, "step": 2180 }, { "epoch": 0.22, "learning_rate": 0.001825896015636759, "loss": 1.7686, "step": 2181 }, { "epoch": 0.22, "learning_rate": 0.0018257159450285206, "loss": 1.7246, "step": 2182 }, { "epoch": 0.22, "learning_rate": 0.0018255357902364885, "loss": 1.6719, "step": 2183 }, { "epoch": 0.22, "learning_rate": 0.0018253555512790305, "loss": 1.835, "step": 2184 }, { "epoch": 0.22, "learning_rate": 0.001825175228174522, "loss": 1.8145, "step": 2185 }, { "epoch": 0.22, "learning_rate": 0.0018249948209413476, "loss": 1.8135, "step": 2186 }, { "epoch": 0.22, "learning_rate": 0.0018248143295979, "loss": 1.7012, "step": 2187 }, { "epoch": 0.22, "learning_rate": 0.0018246337541625813, "loss": 1.9502, "step": 2188 }, { "epoch": 0.22, "learning_rate": 0.001824453094653801, "loss": 1.7812, "step": 2189 }, { "epoch": 0.22, "learning_rate": 0.0018242723510899777, "loss": 1.8594, "step": 2190 }, { "epoch": 0.22, "learning_rate": 0.0018240915234895394, "loss": 1.6758, "step": 2191 }, { "epoch": 0.22, "learning_rate": 0.0018239106118709213, "loss": 1.6807, "step": 2192 }, { "epoch": 0.22, "learning_rate": 0.001823729616252568, "loss": 1.8076, "step": 2193 }, { "epoch": 0.22, "learning_rate": 0.0018235485366529323, "loss": 1.7725, "step": 2194 }, { "epoch": 0.22, "learning_rate": 0.001823367373090476, "loss": 1.708, "step": 2195 }, { "epoch": 0.22, "learning_rate": 0.001823186125583669, "loss": 1.752, "step": 2196 }, { "epoch": 0.22, "learning_rate": 0.00182300479415099, "loss": 1.7373, "step": 2197 }, { "epoch": 0.22, "learning_rate": 0.0018228233788109259, "loss": 1.8125, "step": 2198 }, { "epoch": 0.22, "learning_rate": 0.001822641879581973, "loss": 1.6943, "step": 2199 }, { "epoch": 0.22, "learning_rate": 0.001822460296482635, "loss": 1.6982, "step": 2200 }, { "epoch": 0.22, "learning_rate": 0.0018222786295314253, "loss": 1.7383, "step": 2201 }, { "epoch": 0.22, "learning_rate": 0.0018220968787468648, "loss": 1.7803, "step": 2202 }, { "epoch": 0.22, "learning_rate": 0.001821915044147484, "loss": 1.7881, "step": 2203 }, { "epoch": 0.22, "learning_rate": 0.001821733125751821, "loss": 1.7393, "step": 2204 }, { "epoch": 0.22, "learning_rate": 0.0018215511235784225, "loss": 1.7295, "step": 2205 }, { "epoch": 0.22, "learning_rate": 0.001821369037645845, "loss": 1.8008, "step": 2206 }, { "epoch": 0.22, "learning_rate": 0.0018211868679726515, "loss": 1.8262, "step": 2207 }, { "epoch": 0.22, "learning_rate": 0.0018210046145774156, "loss": 1.7646, "step": 2208 }, { "epoch": 0.22, "learning_rate": 0.0018208222774787183, "loss": 1.7666, "step": 2209 }, { "epoch": 0.22, "learning_rate": 0.001820639856695149, "loss": 1.832, "step": 2210 }, { "epoch": 0.22, "learning_rate": 0.001820457352245306, "loss": 1.7344, "step": 2211 }, { "epoch": 0.22, "learning_rate": 0.0018202747641477962, "loss": 1.7236, "step": 2212 }, { "epoch": 0.22, "learning_rate": 0.0018200920924212349, "loss": 1.8506, "step": 2213 }, { "epoch": 0.22, "learning_rate": 0.0018199093370842461, "loss": 1.791, "step": 2214 }, { "epoch": 0.22, "learning_rate": 0.001819726498155462, "loss": 1.7305, "step": 2215 }, { "epoch": 0.22, "learning_rate": 0.001819543575653523, "loss": 1.8789, "step": 2216 }, { "epoch": 0.22, "learning_rate": 0.0018193605695970795, "loss": 1.6914, "step": 2217 }, { "epoch": 0.22, "learning_rate": 0.0018191774800047887, "loss": 1.7939, "step": 2218 }, { "epoch": 0.22, "learning_rate": 0.001818994306895317, "loss": 1.8037, "step": 2219 }, { "epoch": 0.22, "learning_rate": 0.00181881105028734, "loss": 1.6895, "step": 2220 }, { "epoch": 0.22, "learning_rate": 0.0018186277101995408, "loss": 1.7578, "step": 2221 }, { "epoch": 0.22, "learning_rate": 0.001818444286650611, "loss": 1.8887, "step": 2222 }, { "epoch": 0.22, "learning_rate": 0.0018182607796592514, "loss": 1.8174, "step": 2223 }, { "epoch": 0.22, "learning_rate": 0.0018180771892441713, "loss": 1.7754, "step": 2224 }, { "epoch": 0.22, "learning_rate": 0.0018178935154240879, "loss": 1.7695, "step": 2225 }, { "epoch": 0.22, "learning_rate": 0.0018177097582177271, "loss": 1.7607, "step": 2226 }, { "epoch": 0.22, "learning_rate": 0.0018175259176438235, "loss": 1.8291, "step": 2227 }, { "epoch": 0.22, "learning_rate": 0.00181734199372112, "loss": 1.9072, "step": 2228 }, { "epoch": 0.22, "learning_rate": 0.0018171579864683686, "loss": 1.8506, "step": 2229 }, { "epoch": 0.22, "learning_rate": 0.0018169738959043285, "loss": 1.6689, "step": 2230 }, { "epoch": 0.22, "learning_rate": 0.0018167897220477689, "loss": 1.7324, "step": 2231 }, { "epoch": 0.22, "learning_rate": 0.0018166054649174663, "loss": 1.8271, "step": 2232 }, { "epoch": 0.22, "learning_rate": 0.0018164211245322067, "loss": 1.7012, "step": 2233 }, { "epoch": 0.22, "learning_rate": 0.0018162367009107836, "loss": 1.8047, "step": 2234 }, { "epoch": 0.22, "learning_rate": 0.0018160521940719994, "loss": 1.7324, "step": 2235 }, { "epoch": 0.22, "learning_rate": 0.0018158676040346653, "loss": 1.7998, "step": 2236 }, { "epoch": 0.22, "learning_rate": 0.0018156829308176006, "loss": 1.7441, "step": 2237 }, { "epoch": 0.22, "learning_rate": 0.0018154981744396333, "loss": 1.7275, "step": 2238 }, { "epoch": 0.22, "learning_rate": 0.0018153133349196, "loss": 1.749, "step": 2239 }, { "epoch": 0.22, "learning_rate": 0.0018151284122763446, "loss": 1.667, "step": 2240 }, { "epoch": 0.22, "learning_rate": 0.0018149434065287213, "loss": 1.8975, "step": 2241 }, { "epoch": 0.22, "learning_rate": 0.0018147583176955917, "loss": 1.6816, "step": 2242 }, { "epoch": 0.22, "learning_rate": 0.001814573145795826, "loss": 1.8291, "step": 2243 }, { "epoch": 0.22, "learning_rate": 0.001814387890848303, "loss": 1.835, "step": 2244 }, { "epoch": 0.22, "learning_rate": 0.00181420255287191, "loss": 1.7822, "step": 2245 }, { "epoch": 0.22, "learning_rate": 0.0018140171318855422, "loss": 1.793, "step": 2246 }, { "epoch": 0.22, "learning_rate": 0.001813831627908104, "loss": 1.6963, "step": 2247 }, { "epoch": 0.22, "learning_rate": 0.0018136460409585084, "loss": 1.7109, "step": 2248 }, { "epoch": 0.22, "learning_rate": 0.001813460371055676, "loss": 1.7666, "step": 2249 }, { "epoch": 0.22, "learning_rate": 0.0018132746182185364, "loss": 1.7402, "step": 2250 }, { "epoch": 0.22, "learning_rate": 0.0018130887824660275, "loss": 1.8262, "step": 2251 }, { "epoch": 0.22, "learning_rate": 0.001812902863817096, "loss": 1.6602, "step": 2252 }, { "epoch": 0.22, "learning_rate": 0.0018127168622906963, "loss": 1.9033, "step": 2253 }, { "epoch": 0.22, "learning_rate": 0.0018125307779057922, "loss": 1.7695, "step": 2254 }, { "epoch": 0.22, "learning_rate": 0.0018123446106813553, "loss": 1.792, "step": 2255 }, { "epoch": 0.22, "learning_rate": 0.0018121583606363655, "loss": 1.8574, "step": 2256 }, { "epoch": 0.22, "learning_rate": 0.0018119720277898117, "loss": 1.7021, "step": 2257 }, { "epoch": 0.22, "learning_rate": 0.001811785612160691, "loss": 1.8271, "step": 2258 }, { "epoch": 0.22, "learning_rate": 0.0018115991137680092, "loss": 1.6973, "step": 2259 }, { "epoch": 0.22, "learning_rate": 0.0018114125326307799, "loss": 1.7461, "step": 2260 }, { "epoch": 0.22, "learning_rate": 0.0018112258687680255, "loss": 1.6211, "step": 2261 }, { "epoch": 0.22, "learning_rate": 0.0018110391221987769, "loss": 1.7432, "step": 2262 }, { "epoch": 0.22, "learning_rate": 0.001810852292942074, "loss": 1.7188, "step": 2263 }, { "epoch": 0.22, "learning_rate": 0.0018106653810169636, "loss": 1.7383, "step": 2264 }, { "epoch": 0.22, "learning_rate": 0.0018104783864425018, "loss": 1.7002, "step": 2265 }, { "epoch": 0.22, "learning_rate": 0.001810291309237754, "loss": 1.7949, "step": 2266 }, { "epoch": 0.22, "learning_rate": 0.0018101041494217927, "loss": 1.7617, "step": 2267 }, { "epoch": 0.22, "learning_rate": 0.0018099169070136993, "loss": 1.6748, "step": 2268 }, { "epoch": 0.22, "learning_rate": 0.0018097295820325636, "loss": 1.8662, "step": 2269 }, { "epoch": 0.22, "learning_rate": 0.001809542174497484, "loss": 1.5977, "step": 2270 }, { "epoch": 0.22, "learning_rate": 0.0018093546844275671, "loss": 1.834, "step": 2271 }, { "epoch": 0.22, "learning_rate": 0.0018091671118419283, "loss": 1.707, "step": 2272 }, { "epoch": 0.22, "learning_rate": 0.0018089794567596902, "loss": 1.7061, "step": 2273 }, { "epoch": 0.22, "learning_rate": 0.0018087917191999855, "loss": 1.7861, "step": 2274 }, { "epoch": 0.22, "learning_rate": 0.0018086038991819541, "loss": 1.8174, "step": 2275 }, { "epoch": 0.22, "learning_rate": 0.0018084159967247451, "loss": 1.6436, "step": 2276 }, { "epoch": 0.22, "learning_rate": 0.0018082280118475156, "loss": 1.7666, "step": 2277 }, { "epoch": 0.22, "learning_rate": 0.0018080399445694307, "loss": 1.8467, "step": 2278 }, { "epoch": 0.22, "learning_rate": 0.0018078517949096644, "loss": 1.9736, "step": 2279 }, { "epoch": 0.22, "learning_rate": 0.0018076635628873994, "loss": 1.8145, "step": 2280 }, { "epoch": 0.22, "learning_rate": 0.0018074752485218262, "loss": 1.7842, "step": 2281 }, { "epoch": 0.22, "learning_rate": 0.0018072868518321438, "loss": 1.8896, "step": 2282 }, { "epoch": 0.23, "learning_rate": 0.0018070983728375601, "loss": 1.7031, "step": 2283 }, { "epoch": 0.23, "learning_rate": 0.0018069098115572905, "loss": 1.6963, "step": 2284 }, { "epoch": 0.23, "learning_rate": 0.0018067211680105597, "loss": 1.6426, "step": 2285 }, { "epoch": 0.23, "learning_rate": 0.0018065324422166002, "loss": 1.749, "step": 2286 }, { "epoch": 0.23, "learning_rate": 0.001806343634194653, "loss": 1.7637, "step": 2287 }, { "epoch": 0.23, "learning_rate": 0.0018061547439639676, "loss": 1.7656, "step": 2288 }, { "epoch": 0.23, "learning_rate": 0.0018059657715438018, "loss": 1.7686, "step": 2289 }, { "epoch": 0.23, "learning_rate": 0.001805776716953422, "loss": 1.7461, "step": 2290 }, { "epoch": 0.23, "learning_rate": 0.0018055875802121027, "loss": 1.7285, "step": 2291 }, { "epoch": 0.23, "learning_rate": 0.0018053983613391267, "loss": 1.665, "step": 2292 }, { "epoch": 0.23, "learning_rate": 0.0018052090603537854, "loss": 1.6533, "step": 2293 }, { "epoch": 0.23, "learning_rate": 0.0018050196772753787, "loss": 1.79, "step": 2294 }, { "epoch": 0.23, "learning_rate": 0.0018048302121232145, "loss": 1.7432, "step": 2295 }, { "epoch": 0.23, "learning_rate": 0.0018046406649166091, "loss": 1.6914, "step": 2296 }, { "epoch": 0.23, "learning_rate": 0.0018044510356748878, "loss": 1.6533, "step": 2297 }, { "epoch": 0.23, "learning_rate": 0.001804261324417383, "loss": 1.833, "step": 2298 }, { "epoch": 0.23, "learning_rate": 0.0018040715311634368, "loss": 1.9248, "step": 2299 }, { "epoch": 0.23, "learning_rate": 0.0018038816559323991, "loss": 1.8037, "step": 2300 }, { "epoch": 0.23, "learning_rate": 0.0018036916987436281, "loss": 1.7246, "step": 2301 }, { "epoch": 0.23, "learning_rate": 0.0018035016596164904, "loss": 1.751, "step": 2302 }, { "epoch": 0.23, "learning_rate": 0.0018033115385703608, "loss": 1.8184, "step": 2303 }, { "epoch": 0.23, "learning_rate": 0.0018031213356246225, "loss": 1.7783, "step": 2304 }, { "epoch": 0.23, "learning_rate": 0.0018029310507986676, "loss": 1.7305, "step": 2305 }, { "epoch": 0.23, "learning_rate": 0.0018027406841118958, "loss": 1.6641, "step": 2306 }, { "epoch": 0.23, "learning_rate": 0.0018025502355837157, "loss": 1.8154, "step": 2307 }, { "epoch": 0.23, "learning_rate": 0.0018023597052335436, "loss": 1.8467, "step": 2308 }, { "epoch": 0.23, "learning_rate": 0.001802169093080805, "loss": 1.8291, "step": 2309 }, { "epoch": 0.23, "learning_rate": 0.001801978399144933, "loss": 1.75, "step": 2310 }, { "epoch": 0.23, "learning_rate": 0.0018017876234453693, "loss": 1.6836, "step": 2311 }, { "epoch": 0.23, "learning_rate": 0.0018015967660015642, "loss": 1.7646, "step": 2312 }, { "epoch": 0.23, "learning_rate": 0.0018014058268329758, "loss": 1.6494, "step": 2313 }, { "epoch": 0.23, "learning_rate": 0.001801214805959071, "loss": 1.7051, "step": 2314 }, { "epoch": 0.23, "learning_rate": 0.0018010237033993248, "loss": 1.792, "step": 2315 }, { "epoch": 0.23, "learning_rate": 0.0018008325191732207, "loss": 1.7168, "step": 2316 }, { "epoch": 0.23, "learning_rate": 0.0018006412533002503, "loss": 1.8213, "step": 2317 }, { "epoch": 0.23, "learning_rate": 0.0018004499057999138, "loss": 1.708, "step": 2318 }, { "epoch": 0.23, "learning_rate": 0.001800258476691719, "loss": 1.6396, "step": 2319 }, { "epoch": 0.23, "learning_rate": 0.001800066965995183, "loss": 1.6475, "step": 2320 }, { "epoch": 0.23, "learning_rate": 0.0017998753737298308, "loss": 1.8418, "step": 2321 }, { "epoch": 0.23, "learning_rate": 0.001799683699915196, "loss": 1.8174, "step": 2322 }, { "epoch": 0.23, "learning_rate": 0.0017994919445708196, "loss": 1.7803, "step": 2323 }, { "epoch": 0.23, "learning_rate": 0.0017993001077162517, "loss": 1.8447, "step": 2324 }, { "epoch": 0.23, "learning_rate": 0.0017991081893710507, "loss": 1.7324, "step": 2325 }, { "epoch": 0.23, "learning_rate": 0.0017989161895547833, "loss": 1.6943, "step": 2326 }, { "epoch": 0.23, "learning_rate": 0.0017987241082870243, "loss": 1.7754, "step": 2327 }, { "epoch": 0.23, "learning_rate": 0.0017985319455873566, "loss": 1.7705, "step": 2328 }, { "epoch": 0.23, "learning_rate": 0.0017983397014753714, "loss": 1.6914, "step": 2329 }, { "epoch": 0.23, "learning_rate": 0.0017981473759706693, "loss": 1.8145, "step": 2330 }, { "epoch": 0.23, "learning_rate": 0.0017979549690928579, "loss": 1.7266, "step": 2331 }, { "epoch": 0.23, "learning_rate": 0.0017977624808615533, "loss": 1.7207, "step": 2332 }, { "epoch": 0.23, "learning_rate": 0.0017975699112963808, "loss": 1.8379, "step": 2333 }, { "epoch": 0.23, "learning_rate": 0.0017973772604169729, "loss": 1.6816, "step": 2334 }, { "epoch": 0.23, "learning_rate": 0.001797184528242971, "loss": 1.7871, "step": 2335 }, { "epoch": 0.23, "learning_rate": 0.0017969917147940243, "loss": 1.8076, "step": 2336 }, { "epoch": 0.23, "learning_rate": 0.0017967988200897906, "loss": 1.6094, "step": 2337 }, { "epoch": 0.23, "learning_rate": 0.0017966058441499367, "loss": 1.6348, "step": 2338 }, { "epoch": 0.23, "learning_rate": 0.0017964127869941366, "loss": 1.6777, "step": 2339 }, { "epoch": 0.23, "learning_rate": 0.0017962196486420727, "loss": 1.7598, "step": 2340 }, { "epoch": 0.23, "learning_rate": 0.0017960264291134362, "loss": 1.874, "step": 2341 }, { "epoch": 0.23, "learning_rate": 0.0017958331284279262, "loss": 1.7383, "step": 2342 }, { "epoch": 0.23, "learning_rate": 0.0017956397466052502, "loss": 1.7178, "step": 2343 }, { "epoch": 0.23, "learning_rate": 0.001795446283665124, "loss": 1.6543, "step": 2344 }, { "epoch": 0.23, "learning_rate": 0.0017952527396272715, "loss": 1.9268, "step": 2345 }, { "epoch": 0.23, "learning_rate": 0.001795059114511425, "loss": 1.7852, "step": 2346 }, { "epoch": 0.23, "learning_rate": 0.0017948654083373254, "loss": 1.6074, "step": 2347 }, { "epoch": 0.23, "learning_rate": 0.0017946716211247214, "loss": 1.8857, "step": 2348 }, { "epoch": 0.23, "learning_rate": 0.0017944777528933699, "loss": 1.793, "step": 2349 }, { "epoch": 0.23, "learning_rate": 0.0017942838036630363, "loss": 1.6875, "step": 2350 }, { "epoch": 0.23, "learning_rate": 0.001794089773453494, "loss": 1.8643, "step": 2351 }, { "epoch": 0.23, "learning_rate": 0.0017938956622845252, "loss": 1.7305, "step": 2352 }, { "epoch": 0.23, "learning_rate": 0.00179370147017592, "loss": 1.7891, "step": 2353 }, { "epoch": 0.23, "learning_rate": 0.0017935071971474765, "loss": 1.707, "step": 2354 }, { "epoch": 0.23, "learning_rate": 0.0017933128432190018, "loss": 1.8184, "step": 2355 }, { "epoch": 0.23, "learning_rate": 0.0017931184084103104, "loss": 1.7656, "step": 2356 }, { "epoch": 0.23, "learning_rate": 0.0017929238927412254, "loss": 1.8408, "step": 2357 }, { "epoch": 0.23, "learning_rate": 0.001792729296231578, "loss": 1.7695, "step": 2358 }, { "epoch": 0.23, "learning_rate": 0.0017925346189012084, "loss": 1.7832, "step": 2359 }, { "epoch": 0.23, "learning_rate": 0.001792339860769964, "loss": 1.5938, "step": 2360 }, { "epoch": 0.23, "learning_rate": 0.0017921450218577007, "loss": 1.6953, "step": 2361 }, { "epoch": 0.23, "learning_rate": 0.0017919501021842832, "loss": 1.6924, "step": 2362 }, { "epoch": 0.23, "learning_rate": 0.0017917551017695838, "loss": 1.7842, "step": 2363 }, { "epoch": 0.23, "learning_rate": 0.0017915600206334834, "loss": 1.7881, "step": 2364 }, { "epoch": 0.23, "learning_rate": 0.0017913648587958712, "loss": 1.8564, "step": 2365 }, { "epoch": 0.23, "learning_rate": 0.001791169616276644, "loss": 1.752, "step": 2366 }, { "epoch": 0.23, "learning_rate": 0.0017909742930957075, "loss": 1.7051, "step": 2367 }, { "epoch": 0.23, "learning_rate": 0.0017907788892729756, "loss": 1.6904, "step": 2368 }, { "epoch": 0.23, "learning_rate": 0.0017905834048283697, "loss": 1.6953, "step": 2369 }, { "epoch": 0.23, "learning_rate": 0.0017903878397818203, "loss": 1.7842, "step": 2370 }, { "epoch": 0.23, "learning_rate": 0.0017901921941532654, "loss": 1.7715, "step": 2371 }, { "epoch": 0.23, "learning_rate": 0.0017899964679626522, "loss": 1.7607, "step": 2372 }, { "epoch": 0.23, "learning_rate": 0.001789800661229935, "loss": 1.8164, "step": 2373 }, { "epoch": 0.23, "learning_rate": 0.0017896047739750765, "loss": 1.6367, "step": 2374 }, { "epoch": 0.23, "learning_rate": 0.0017894088062180482, "loss": 1.7793, "step": 2375 }, { "epoch": 0.23, "learning_rate": 0.00178921275797883, "loss": 1.7168, "step": 2376 }, { "epoch": 0.23, "learning_rate": 0.0017890166292774089, "loss": 1.8564, "step": 2377 }, { "epoch": 0.23, "learning_rate": 0.0017888204201337806, "loss": 1.6562, "step": 2378 }, { "epoch": 0.23, "learning_rate": 0.0017886241305679496, "loss": 1.7324, "step": 2379 }, { "epoch": 0.23, "learning_rate": 0.0017884277605999279, "loss": 1.8477, "step": 2380 }, { "epoch": 0.23, "learning_rate": 0.0017882313102497358, "loss": 1.6904, "step": 2381 }, { "epoch": 0.23, "learning_rate": 0.0017880347795374018, "loss": 1.7559, "step": 2382 }, { "epoch": 0.23, "learning_rate": 0.001787838168482963, "loss": 1.748, "step": 2383 }, { "epoch": 0.24, "learning_rate": 0.0017876414771064643, "loss": 1.6709, "step": 2384 }, { "epoch": 0.24, "learning_rate": 0.001787444705427959, "loss": 1.7676, "step": 2385 }, { "epoch": 0.24, "learning_rate": 0.001787247853467508, "loss": 1.7695, "step": 2386 }, { "epoch": 0.24, "learning_rate": 0.0017870509212451816, "loss": 1.6436, "step": 2387 }, { "epoch": 0.24, "learning_rate": 0.0017868539087810564, "loss": 1.749, "step": 2388 }, { "epoch": 0.24, "learning_rate": 0.0017866568160952196, "loss": 1.7539, "step": 2389 }, { "epoch": 0.24, "learning_rate": 0.0017864596432077644, "loss": 1.7441, "step": 2390 }, { "epoch": 0.24, "learning_rate": 0.0017862623901387933, "loss": 1.7266, "step": 2391 }, { "epoch": 0.24, "learning_rate": 0.0017860650569084169, "loss": 1.751, "step": 2392 }, { "epoch": 0.24, "learning_rate": 0.0017858676435367538, "loss": 1.7422, "step": 2393 }, { "epoch": 0.24, "learning_rate": 0.0017856701500439302, "loss": 1.7178, "step": 2394 }, { "epoch": 0.24, "learning_rate": 0.001785472576450082, "loss": 1.752, "step": 2395 }, { "epoch": 0.24, "learning_rate": 0.0017852749227753517, "loss": 1.8115, "step": 2396 }, { "epoch": 0.24, "learning_rate": 0.0017850771890398907, "loss": 1.5303, "step": 2397 }, { "epoch": 0.24, "learning_rate": 0.0017848793752638585, "loss": 1.7197, "step": 2398 }, { "epoch": 0.24, "learning_rate": 0.0017846814814674228, "loss": 1.6953, "step": 2399 }, { "epoch": 0.24, "learning_rate": 0.001784483507670759, "loss": 1.709, "step": 2400 }, { "epoch": 0.24, "learning_rate": 0.0017842854538940515, "loss": 1.79, "step": 2401 }, { "epoch": 0.24, "learning_rate": 0.001784087320157492, "loss": 1.709, "step": 2402 }, { "epoch": 0.24, "learning_rate": 0.001783889106481281, "loss": 1.7471, "step": 2403 }, { "epoch": 0.24, "learning_rate": 0.0017836908128856268, "loss": 1.7656, "step": 2404 }, { "epoch": 0.24, "learning_rate": 0.0017834924393907458, "loss": 1.7803, "step": 2405 }, { "epoch": 0.24, "learning_rate": 0.0017832939860168626, "loss": 1.6729, "step": 2406 }, { "epoch": 0.24, "learning_rate": 0.0017830954527842105, "loss": 1.7041, "step": 2407 }, { "epoch": 0.24, "learning_rate": 0.0017828968397130298, "loss": 1.7988, "step": 2408 }, { "epoch": 0.24, "learning_rate": 0.00178269814682357, "loss": 1.7588, "step": 2409 }, { "epoch": 0.24, "learning_rate": 0.0017824993741360883, "loss": 1.7207, "step": 2410 }, { "epoch": 0.24, "learning_rate": 0.0017823005216708499, "loss": 1.8691, "step": 2411 }, { "epoch": 0.24, "learning_rate": 0.0017821015894481285, "loss": 1.7852, "step": 2412 }, { "epoch": 0.24, "learning_rate": 0.0017819025774882058, "loss": 1.6748, "step": 2413 }, { "epoch": 0.24, "learning_rate": 0.001781703485811371, "loss": 1.6943, "step": 2414 }, { "epoch": 0.24, "learning_rate": 0.001781504314437923, "loss": 1.7617, "step": 2415 }, { "epoch": 0.24, "learning_rate": 0.0017813050633881668, "loss": 1.6738, "step": 2416 }, { "epoch": 0.24, "learning_rate": 0.0017811057326824172, "loss": 1.5859, "step": 2417 }, { "epoch": 0.24, "learning_rate": 0.001780906322340996, "loss": 1.6416, "step": 2418 }, { "epoch": 0.24, "learning_rate": 0.001780706832384234, "loss": 1.752, "step": 2419 }, { "epoch": 0.24, "learning_rate": 0.0017805072628324695, "loss": 1.707, "step": 2420 }, { "epoch": 0.24, "learning_rate": 0.001780307613706049, "loss": 1.6953, "step": 2421 }, { "epoch": 0.24, "learning_rate": 0.0017801078850253273, "loss": 1.8213, "step": 2422 }, { "epoch": 0.24, "learning_rate": 0.0017799080768106674, "loss": 1.7217, "step": 2423 }, { "epoch": 0.24, "learning_rate": 0.0017797081890824403, "loss": 1.6221, "step": 2424 }, { "epoch": 0.24, "learning_rate": 0.0017795082218610245, "loss": 1.833, "step": 2425 }, { "epoch": 0.24, "learning_rate": 0.0017793081751668076, "loss": 1.7412, "step": 2426 }, { "epoch": 0.24, "learning_rate": 0.001779108049020185, "loss": 1.6992, "step": 2427 }, { "epoch": 0.24, "learning_rate": 0.0017789078434415598, "loss": 1.7617, "step": 2428 }, { "epoch": 0.24, "learning_rate": 0.0017787075584513431, "loss": 1.7764, "step": 2429 }, { "epoch": 0.24, "learning_rate": 0.0017785071940699556, "loss": 1.6865, "step": 2430 }, { "epoch": 0.24, "learning_rate": 0.0017783067503178236, "loss": 1.8018, "step": 2431 }, { "epoch": 0.24, "learning_rate": 0.0017781062272153837, "loss": 1.749, "step": 2432 }, { "epoch": 0.24, "learning_rate": 0.0017779056247830794, "loss": 1.792, "step": 2433 }, { "epoch": 0.24, "learning_rate": 0.0017777049430413626, "loss": 1.7188, "step": 2434 }, { "epoch": 0.24, "learning_rate": 0.0017775041820106937, "loss": 1.8613, "step": 2435 }, { "epoch": 0.24, "learning_rate": 0.00177730334171154, "loss": 1.7119, "step": 2436 }, { "epoch": 0.24, "learning_rate": 0.0017771024221643783, "loss": 1.8486, "step": 2437 }, { "epoch": 0.24, "learning_rate": 0.0017769014233896927, "loss": 1.7607, "step": 2438 }, { "epoch": 0.24, "learning_rate": 0.0017767003454079759, "loss": 1.8193, "step": 2439 }, { "epoch": 0.24, "learning_rate": 0.0017764991882397276, "loss": 1.6689, "step": 2440 }, { "epoch": 0.24, "learning_rate": 0.0017762979519054566, "loss": 1.7266, "step": 2441 }, { "epoch": 0.24, "learning_rate": 0.001776096636425679, "loss": 1.6699, "step": 2442 }, { "epoch": 0.24, "learning_rate": 0.0017758952418209205, "loss": 1.7881, "step": 2443 }, { "epoch": 0.24, "learning_rate": 0.001775693768111713, "loss": 1.7236, "step": 2444 }, { "epoch": 0.24, "learning_rate": 0.0017754922153185972, "loss": 1.8076, "step": 2445 }, { "epoch": 0.24, "learning_rate": 0.0017752905834621223, "loss": 1.7793, "step": 2446 }, { "epoch": 0.24, "learning_rate": 0.0017750888725628446, "loss": 1.6768, "step": 2447 }, { "epoch": 0.24, "learning_rate": 0.0017748870826413295, "loss": 1.7969, "step": 2448 }, { "epoch": 0.24, "learning_rate": 0.0017746852137181502, "loss": 1.6729, "step": 2449 }, { "epoch": 0.24, "learning_rate": 0.001774483265813887, "loss": 1.7217, "step": 2450 }, { "epoch": 0.24, "learning_rate": 0.0017742812389491295, "loss": 1.7334, "step": 2451 }, { "epoch": 0.24, "learning_rate": 0.001774079133144475, "loss": 1.7588, "step": 2452 }, { "epoch": 0.24, "learning_rate": 0.0017738769484205278, "loss": 1.7148, "step": 2453 }, { "epoch": 0.24, "learning_rate": 0.0017736746847979025, "loss": 1.7305, "step": 2454 }, { "epoch": 0.24, "learning_rate": 0.001773472342297219, "loss": 1.8135, "step": 2455 }, { "epoch": 0.24, "learning_rate": 0.0017732699209391076, "loss": 1.748, "step": 2456 }, { "epoch": 0.24, "learning_rate": 0.0017730674207442052, "loss": 1.7705, "step": 2457 }, { "epoch": 0.24, "learning_rate": 0.0017728648417331574, "loss": 1.709, "step": 2458 }, { "epoch": 0.24, "learning_rate": 0.0017726621839266176, "loss": 1.7891, "step": 2459 }, { "epoch": 0.24, "learning_rate": 0.001772459447345247, "loss": 1.8457, "step": 2460 }, { "epoch": 0.24, "learning_rate": 0.0017722566320097154, "loss": 1.7441, "step": 2461 }, { "epoch": 0.24, "learning_rate": 0.0017720537379407003, "loss": 1.7842, "step": 2462 }, { "epoch": 0.24, "learning_rate": 0.0017718507651588874, "loss": 1.7979, "step": 2463 }, { "epoch": 0.24, "learning_rate": 0.0017716477136849697, "loss": 1.8027, "step": 2464 }, { "epoch": 0.24, "learning_rate": 0.0017714445835396494, "loss": 1.8066, "step": 2465 }, { "epoch": 0.24, "learning_rate": 0.001771241374743636, "loss": 1.7012, "step": 2466 }, { "epoch": 0.24, "learning_rate": 0.0017710380873176468, "loss": 1.7578, "step": 2467 }, { "epoch": 0.24, "learning_rate": 0.001770834721282408, "loss": 1.7383, "step": 2468 }, { "epoch": 0.24, "learning_rate": 0.0017706312766586523, "loss": 1.6445, "step": 2469 }, { "epoch": 0.24, "learning_rate": 0.0017704277534671227, "loss": 1.7832, "step": 2470 }, { "epoch": 0.24, "learning_rate": 0.0017702241517285682, "loss": 1.7744, "step": 2471 }, { "epoch": 0.24, "learning_rate": 0.0017700204714637465, "loss": 1.71, "step": 2472 }, { "epoch": 0.24, "learning_rate": 0.001769816712693423, "loss": 1.6768, "step": 2473 }, { "epoch": 0.24, "learning_rate": 0.0017696128754383722, "loss": 1.8594, "step": 2474 }, { "epoch": 0.24, "learning_rate": 0.0017694089597193755, "loss": 1.6797, "step": 2475 }, { "epoch": 0.24, "learning_rate": 0.0017692049655572222, "loss": 1.6865, "step": 2476 }, { "epoch": 0.24, "learning_rate": 0.0017690008929727105, "loss": 1.666, "step": 2477 }, { "epoch": 0.24, "learning_rate": 0.0017687967419866457, "loss": 1.8027, "step": 2478 }, { "epoch": 0.24, "learning_rate": 0.001768592512619842, "loss": 1.5732, "step": 2479 }, { "epoch": 0.24, "learning_rate": 0.0017683882048931206, "loss": 1.7627, "step": 2480 }, { "epoch": 0.24, "learning_rate": 0.0017681838188273115, "loss": 1.7646, "step": 2481 }, { "epoch": 0.24, "learning_rate": 0.0017679793544432526, "loss": 1.6035, "step": 2482 }, { "epoch": 0.24, "learning_rate": 0.0017677748117617886, "loss": 1.7002, "step": 2483 }, { "epoch": 0.24, "learning_rate": 0.0017675701908037742, "loss": 1.8281, "step": 2484 }, { "epoch": 0.24, "learning_rate": 0.0017673654915900704, "loss": 1.7168, "step": 2485 }, { "epoch": 0.25, "learning_rate": 0.0017671607141415472, "loss": 1.7422, "step": 2486 }, { "epoch": 0.25, "learning_rate": 0.0017669558584790816, "loss": 1.75, "step": 2487 }, { "epoch": 0.25, "learning_rate": 0.0017667509246235597, "loss": 1.7441, "step": 2488 }, { "epoch": 0.25, "learning_rate": 0.0017665459125958747, "loss": 1.7549, "step": 2489 }, { "epoch": 0.25, "learning_rate": 0.0017663408224169283, "loss": 1.833, "step": 2490 }, { "epoch": 0.25, "learning_rate": 0.0017661356541076297, "loss": 1.7432, "step": 2491 }, { "epoch": 0.25, "learning_rate": 0.0017659304076888967, "loss": 1.8018, "step": 2492 }, { "epoch": 0.25, "learning_rate": 0.0017657250831816542, "loss": 1.916, "step": 2493 }, { "epoch": 0.25, "learning_rate": 0.0017655196806068355, "loss": 1.834, "step": 2494 }, { "epoch": 0.25, "learning_rate": 0.001765314199985383, "loss": 1.6611, "step": 2495 }, { "epoch": 0.25, "learning_rate": 0.0017651086413382445, "loss": 1.8311, "step": 2496 }, { "epoch": 0.25, "learning_rate": 0.001764903004686378, "loss": 1.6777, "step": 2497 }, { "epoch": 0.25, "learning_rate": 0.0017646972900507488, "loss": 1.8301, "step": 2498 }, { "epoch": 0.25, "learning_rate": 0.0017644914974523293, "loss": 1.8096, "step": 2499 }, { "epoch": 0.25, "learning_rate": 0.0017642856269121012, "loss": 1.75, "step": 2500 }, { "epoch": 0.25, "learning_rate": 0.0017640796784510536, "loss": 1.7842, "step": 2501 }, { "epoch": 0.25, "learning_rate": 0.0017638736520901829, "loss": 1.7598, "step": 2502 }, { "epoch": 0.25, "learning_rate": 0.0017636675478504941, "loss": 1.5703, "step": 2503 }, { "epoch": 0.25, "learning_rate": 0.0017634613657530008, "loss": 1.6572, "step": 2504 }, { "epoch": 0.25, "learning_rate": 0.0017632551058187227, "loss": 1.7861, "step": 2505 }, { "epoch": 0.25, "learning_rate": 0.001763048768068689, "loss": 1.6309, "step": 2506 }, { "epoch": 0.25, "learning_rate": 0.0017628423525239368, "loss": 1.791, "step": 2507 }, { "epoch": 0.25, "learning_rate": 0.00176263585920551, "loss": 1.7773, "step": 2508 }, { "epoch": 0.25, "learning_rate": 0.0017624292881344612, "loss": 1.7363, "step": 2509 }, { "epoch": 0.25, "learning_rate": 0.0017622226393318513, "loss": 1.6768, "step": 2510 }, { "epoch": 0.25, "learning_rate": 0.0017620159128187481, "loss": 1.7568, "step": 2511 }, { "epoch": 0.25, "learning_rate": 0.0017618091086162282, "loss": 1.7998, "step": 2512 }, { "epoch": 0.25, "learning_rate": 0.0017616022267453756, "loss": 1.71, "step": 2513 }, { "epoch": 0.25, "learning_rate": 0.0017613952672272829, "loss": 1.667, "step": 2514 }, { "epoch": 0.25, "learning_rate": 0.0017611882300830493, "loss": 1.7129, "step": 2515 }, { "epoch": 0.25, "learning_rate": 0.0017609811153337835, "loss": 1.7246, "step": 2516 }, { "epoch": 0.25, "learning_rate": 0.0017607739230006013, "loss": 1.7891, "step": 2517 }, { "epoch": 0.25, "learning_rate": 0.001760566653104626, "loss": 1.793, "step": 2518 }, { "epoch": 0.25, "learning_rate": 0.0017603593056669897, "loss": 1.7373, "step": 2519 }, { "epoch": 0.25, "learning_rate": 0.0017601518807088318, "loss": 1.6377, "step": 2520 }, { "epoch": 0.25, "learning_rate": 0.0017599443782512997, "loss": 1.7744, "step": 2521 }, { "epoch": 0.25, "learning_rate": 0.0017597367983155491, "loss": 1.6143, "step": 2522 }, { "epoch": 0.25, "learning_rate": 0.001759529140922743, "loss": 1.7529, "step": 2523 }, { "epoch": 0.25, "learning_rate": 0.0017593214060940526, "loss": 1.7412, "step": 2524 }, { "epoch": 0.25, "learning_rate": 0.0017591135938506573, "loss": 1.7227, "step": 2525 }, { "epoch": 0.25, "learning_rate": 0.0017589057042137438, "loss": 1.6602, "step": 2526 }, { "epoch": 0.25, "learning_rate": 0.001758697737204507, "loss": 1.7021, "step": 2527 }, { "epoch": 0.25, "learning_rate": 0.0017584896928441497, "loss": 1.6621, "step": 2528 }, { "epoch": 0.25, "learning_rate": 0.0017582815711538822, "loss": 1.7197, "step": 2529 }, { "epoch": 0.25, "learning_rate": 0.0017580733721549238, "loss": 1.6924, "step": 2530 }, { "epoch": 0.25, "learning_rate": 0.0017578650958685002, "loss": 1.7754, "step": 2531 }, { "epoch": 0.25, "learning_rate": 0.001757656742315846, "loss": 1.7715, "step": 2532 }, { "epoch": 0.25, "learning_rate": 0.0017574483115182034, "loss": 1.7852, "step": 2533 }, { "epoch": 0.25, "learning_rate": 0.0017572398034968223, "loss": 1.8125, "step": 2534 }, { "epoch": 0.25, "learning_rate": 0.0017570312182729607, "loss": 1.7119, "step": 2535 }, { "epoch": 0.25, "learning_rate": 0.001756822555867884, "loss": 1.8066, "step": 2536 }, { "epoch": 0.25, "learning_rate": 0.0017566138163028664, "loss": 1.7471, "step": 2537 }, { "epoch": 0.25, "learning_rate": 0.001756404999599189, "loss": 1.7666, "step": 2538 }, { "epoch": 0.25, "learning_rate": 0.0017561961057781417, "loss": 1.7754, "step": 2539 }, { "epoch": 0.25, "learning_rate": 0.0017559871348610213, "loss": 1.6523, "step": 2540 }, { "epoch": 0.25, "learning_rate": 0.0017557780868691331, "loss": 1.6387, "step": 2541 }, { "epoch": 0.25, "learning_rate": 0.0017555689618237898, "loss": 1.7295, "step": 2542 }, { "epoch": 0.25, "learning_rate": 0.0017553597597463125, "loss": 1.7422, "step": 2543 }, { "epoch": 0.25, "learning_rate": 0.0017551504806580297, "loss": 1.7832, "step": 2544 }, { "epoch": 0.25, "learning_rate": 0.001754941124580278, "loss": 1.8428, "step": 2545 }, { "epoch": 0.25, "learning_rate": 0.0017547316915344018, "loss": 1.7656, "step": 2546 }, { "epoch": 0.25, "learning_rate": 0.0017545221815417533, "loss": 1.7148, "step": 2547 }, { "epoch": 0.25, "learning_rate": 0.0017543125946236926, "loss": 1.7744, "step": 2548 }, { "epoch": 0.25, "learning_rate": 0.0017541029308015877, "loss": 1.7041, "step": 2549 }, { "epoch": 0.25, "learning_rate": 0.001753893190096814, "loss": 1.7734, "step": 2550 }, { "epoch": 0.25, "learning_rate": 0.0017536833725307554, "loss": 1.8291, "step": 2551 }, { "epoch": 0.25, "learning_rate": 0.0017534734781248032, "loss": 1.6855, "step": 2552 }, { "epoch": 0.25, "learning_rate": 0.0017532635069003567, "loss": 1.7607, "step": 2553 }, { "epoch": 0.25, "learning_rate": 0.001753053458878823, "loss": 1.7715, "step": 2554 }, { "epoch": 0.25, "learning_rate": 0.001752843334081617, "loss": 1.6924, "step": 2555 }, { "epoch": 0.25, "learning_rate": 0.0017526331325301613, "loss": 1.7568, "step": 2556 }, { "epoch": 0.25, "learning_rate": 0.0017524228542458867, "loss": 1.7363, "step": 2557 }, { "epoch": 0.25, "learning_rate": 0.0017522124992502314, "loss": 1.7441, "step": 2558 }, { "epoch": 0.25, "learning_rate": 0.0017520020675646415, "loss": 1.7275, "step": 2559 }, { "epoch": 0.25, "learning_rate": 0.0017517915592105715, "loss": 1.6621, "step": 2560 }, { "epoch": 0.25, "learning_rate": 0.001751580974209483, "loss": 1.6836, "step": 2561 }, { "epoch": 0.25, "learning_rate": 0.001751370312582846, "loss": 1.7822, "step": 2562 }, { "epoch": 0.25, "learning_rate": 0.0017511595743521372, "loss": 1.7139, "step": 2563 }, { "epoch": 0.25, "learning_rate": 0.0017509487595388422, "loss": 1.75, "step": 2564 }, { "epoch": 0.25, "learning_rate": 0.0017507378681644546, "loss": 1.6533, "step": 2565 }, { "epoch": 0.25, "learning_rate": 0.0017505269002504747, "loss": 1.7275, "step": 2566 }, { "epoch": 0.25, "learning_rate": 0.0017503158558184113, "loss": 1.7998, "step": 2567 }, { "epoch": 0.25, "learning_rate": 0.0017501047348897813, "loss": 1.7832, "step": 2568 }, { "epoch": 0.25, "learning_rate": 0.0017498935374861088, "loss": 1.666, "step": 2569 }, { "epoch": 0.25, "learning_rate": 0.0017496822636289253, "loss": 1.6953, "step": 2570 }, { "epoch": 0.25, "learning_rate": 0.0017494709133397718, "loss": 1.7676, "step": 2571 }, { "epoch": 0.25, "learning_rate": 0.0017492594866401951, "loss": 1.6953, "step": 2572 }, { "epoch": 0.25, "learning_rate": 0.001749047983551751, "loss": 1.75, "step": 2573 }, { "epoch": 0.25, "learning_rate": 0.0017488364040960027, "loss": 1.7529, "step": 2574 }, { "epoch": 0.25, "learning_rate": 0.0017486247482945215, "loss": 1.8008, "step": 2575 }, { "epoch": 0.25, "learning_rate": 0.001748413016168886, "loss": 1.8379, "step": 2576 }, { "epoch": 0.25, "learning_rate": 0.0017482012077406827, "loss": 1.583, "step": 2577 }, { "epoch": 0.25, "learning_rate": 0.0017479893230315062, "loss": 1.7627, "step": 2578 }, { "epoch": 0.25, "learning_rate": 0.0017477773620629589, "loss": 1.8115, "step": 2579 }, { "epoch": 0.25, "learning_rate": 0.00174756532485665, "loss": 1.709, "step": 2580 }, { "epoch": 0.25, "learning_rate": 0.0017473532114341982, "loss": 1.8525, "step": 2581 }, { "epoch": 0.25, "learning_rate": 0.0017471410218172282, "loss": 1.8467, "step": 2582 }, { "epoch": 0.25, "learning_rate": 0.0017469287560273735, "loss": 1.9209, "step": 2583 }, { "epoch": 0.25, "learning_rate": 0.0017467164140862752, "loss": 1.8291, "step": 2584 }, { "epoch": 0.25, "learning_rate": 0.0017465039960155822, "loss": 1.6543, "step": 2585 }, { "epoch": 0.25, "learning_rate": 0.0017462915018369505, "loss": 1.79, "step": 2586 }, { "epoch": 0.26, "learning_rate": 0.001746078931572045, "loss": 1.7314, "step": 2587 }, { "epoch": 0.26, "learning_rate": 0.001745866285242538, "loss": 1.667, "step": 2588 }, { "epoch": 0.26, "learning_rate": 0.0017456535628701086, "loss": 1.6357, "step": 2589 }, { "epoch": 0.26, "learning_rate": 0.0017454407644764448, "loss": 1.7168, "step": 2590 }, { "epoch": 0.26, "learning_rate": 0.0017452278900832417, "loss": 1.75, "step": 2591 }, { "epoch": 0.26, "learning_rate": 0.0017450149397122024, "loss": 1.6826, "step": 2592 }, { "epoch": 0.26, "learning_rate": 0.0017448019133850378, "loss": 1.6689, "step": 2593 }, { "epoch": 0.26, "learning_rate": 0.0017445888111234664, "loss": 1.6465, "step": 2594 }, { "epoch": 0.26, "learning_rate": 0.0017443756329492146, "loss": 1.6406, "step": 2595 }, { "epoch": 0.26, "learning_rate": 0.0017441623788840169, "loss": 1.7188, "step": 2596 }, { "epoch": 0.26, "learning_rate": 0.001743949048949614, "loss": 1.7812, "step": 2597 }, { "epoch": 0.26, "learning_rate": 0.0017437356431677562, "loss": 1.7227, "step": 2598 }, { "epoch": 0.26, "learning_rate": 0.0017435221615602005, "loss": 1.7314, "step": 2599 }, { "epoch": 0.26, "learning_rate": 0.001743308604148712, "loss": 1.7812, "step": 2600 }, { "epoch": 0.26, "learning_rate": 0.0017430949709550637, "loss": 1.7314, "step": 2601 }, { "epoch": 0.26, "learning_rate": 0.0017428812620010352, "loss": 1.6055, "step": 2602 }, { "epoch": 0.26, "learning_rate": 0.0017426674773084153, "loss": 1.7559, "step": 2603 }, { "epoch": 0.26, "learning_rate": 0.0017424536168989997, "loss": 1.7354, "step": 2604 }, { "epoch": 0.26, "learning_rate": 0.001742239680794592, "loss": 1.7773, "step": 2605 }, { "epoch": 0.26, "learning_rate": 0.0017420256690170036, "loss": 1.7461, "step": 2606 }, { "epoch": 0.26, "learning_rate": 0.0017418115815880536, "loss": 1.8145, "step": 2607 }, { "epoch": 0.26, "learning_rate": 0.0017415974185295685, "loss": 1.6582, "step": 2608 }, { "epoch": 0.26, "learning_rate": 0.0017413831798633829, "loss": 1.7246, "step": 2609 }, { "epoch": 0.26, "learning_rate": 0.0017411688656113389, "loss": 1.6875, "step": 2610 }, { "epoch": 0.26, "learning_rate": 0.0017409544757952863, "loss": 1.6709, "step": 2611 }, { "epoch": 0.26, "learning_rate": 0.001740740010437083, "loss": 1.8086, "step": 2612 }, { "epoch": 0.26, "learning_rate": 0.0017405254695585942, "loss": 1.7188, "step": 2613 }, { "epoch": 0.26, "learning_rate": 0.0017403108531816925, "loss": 1.7217, "step": 2614 }, { "epoch": 0.26, "learning_rate": 0.0017400961613282587, "loss": 1.5059, "step": 2615 }, { "epoch": 0.26, "learning_rate": 0.0017398813940201815, "loss": 1.7188, "step": 2616 }, { "epoch": 0.26, "learning_rate": 0.0017396665512793565, "loss": 1.7373, "step": 2617 }, { "epoch": 0.26, "learning_rate": 0.001739451633127688, "loss": 1.6182, "step": 2618 }, { "epoch": 0.26, "learning_rate": 0.001739236639587087, "loss": 1.7158, "step": 2619 }, { "epoch": 0.26, "learning_rate": 0.0017390215706794727, "loss": 1.6846, "step": 2620 }, { "epoch": 0.26, "learning_rate": 0.0017388064264267722, "loss": 1.7725, "step": 2621 }, { "epoch": 0.26, "learning_rate": 0.0017385912068509195, "loss": 1.7354, "step": 2622 }, { "epoch": 0.26, "learning_rate": 0.0017383759119738574, "loss": 1.8242, "step": 2623 }, { "epoch": 0.26, "learning_rate": 0.001738160541817535, "loss": 1.7627, "step": 2624 }, { "epoch": 0.26, "learning_rate": 0.0017379450964039103, "loss": 1.6143, "step": 2625 }, { "epoch": 0.26, "learning_rate": 0.0017377295757549484, "loss": 1.8799, "step": 2626 }, { "epoch": 0.26, "learning_rate": 0.0017375139798926223, "loss": 1.8379, "step": 2627 }, { "epoch": 0.26, "learning_rate": 0.0017372983088389123, "loss": 1.7383, "step": 2628 }, { "epoch": 0.26, "learning_rate": 0.0017370825626158064, "loss": 1.7461, "step": 2629 }, { "epoch": 0.26, "learning_rate": 0.0017368667412453011, "loss": 1.8066, "step": 2630 }, { "epoch": 0.26, "learning_rate": 0.0017366508447493997, "loss": 1.7588, "step": 2631 }, { "epoch": 0.26, "learning_rate": 0.0017364348731501132, "loss": 1.7725, "step": 2632 }, { "epoch": 0.26, "learning_rate": 0.0017362188264694605, "loss": 1.7852, "step": 2633 }, { "epoch": 0.26, "learning_rate": 0.0017360027047294679, "loss": 1.6572, "step": 2634 }, { "epoch": 0.26, "learning_rate": 0.0017357865079521701, "loss": 1.6738, "step": 2635 }, { "epoch": 0.26, "learning_rate": 0.0017355702361596084, "loss": 1.7637, "step": 2636 }, { "epoch": 0.26, "learning_rate": 0.0017353538893738324, "loss": 1.749, "step": 2637 }, { "epoch": 0.26, "learning_rate": 0.0017351374676168989, "loss": 1.5918, "step": 2638 }, { "epoch": 0.26, "learning_rate": 0.0017349209709108732, "loss": 1.7451, "step": 2639 }, { "epoch": 0.26, "learning_rate": 0.0017347043992778275, "loss": 1.6396, "step": 2640 }, { "epoch": 0.26, "learning_rate": 0.0017344877527398416, "loss": 1.6738, "step": 2641 }, { "epoch": 0.26, "learning_rate": 0.0017342710313190035, "loss": 1.7217, "step": 2642 }, { "epoch": 0.26, "learning_rate": 0.0017340542350374082, "loss": 1.79, "step": 2643 }, { "epoch": 0.26, "learning_rate": 0.0017338373639171586, "loss": 1.709, "step": 2644 }, { "epoch": 0.26, "learning_rate": 0.0017336204179803651, "loss": 1.6123, "step": 2645 }, { "epoch": 0.26, "learning_rate": 0.0017334033972491465, "loss": 1.6084, "step": 2646 }, { "epoch": 0.26, "learning_rate": 0.0017331863017456279, "loss": 1.7969, "step": 2647 }, { "epoch": 0.26, "learning_rate": 0.0017329691314919434, "loss": 1.6729, "step": 2648 }, { "epoch": 0.26, "learning_rate": 0.0017327518865102335, "loss": 1.6631, "step": 2649 }, { "epoch": 0.26, "learning_rate": 0.0017325345668226472, "loss": 1.7734, "step": 2650 }, { "epoch": 0.26, "learning_rate": 0.0017323171724513405, "loss": 1.8369, "step": 2651 }, { "epoch": 0.26, "learning_rate": 0.0017320997034184775, "loss": 1.7334, "step": 2652 }, { "epoch": 0.26, "learning_rate": 0.0017318821597462294, "loss": 1.75, "step": 2653 }, { "epoch": 0.26, "learning_rate": 0.0017316645414567758, "loss": 1.7529, "step": 2654 }, { "epoch": 0.26, "learning_rate": 0.0017314468485723031, "loss": 1.7432, "step": 2655 }, { "epoch": 0.26, "learning_rate": 0.001731229081115006, "loss": 1.9141, "step": 2656 }, { "epoch": 0.26, "learning_rate": 0.0017310112391070859, "loss": 1.7275, "step": 2657 }, { "epoch": 0.26, "learning_rate": 0.0017307933225707527, "loss": 1.7549, "step": 2658 }, { "epoch": 0.26, "learning_rate": 0.001730575331528223, "loss": 1.7295, "step": 2659 }, { "epoch": 0.26, "learning_rate": 0.0017303572660017227, "loss": 1.7549, "step": 2660 }, { "epoch": 0.26, "learning_rate": 0.0017301391260134832, "loss": 1.7188, "step": 2661 }, { "epoch": 0.26, "learning_rate": 0.0017299209115857442, "loss": 1.6973, "step": 2662 }, { "epoch": 0.26, "learning_rate": 0.0017297026227407538, "loss": 1.7578, "step": 2663 }, { "epoch": 0.26, "learning_rate": 0.0017294842595007667, "loss": 1.6758, "step": 2664 }, { "epoch": 0.26, "learning_rate": 0.001729265821888046, "loss": 1.7412, "step": 2665 }, { "epoch": 0.26, "learning_rate": 0.0017290473099248614, "loss": 1.709, "step": 2666 }, { "epoch": 0.26, "learning_rate": 0.0017288287236334914, "loss": 1.6289, "step": 2667 }, { "epoch": 0.26, "learning_rate": 0.001728610063036221, "loss": 1.7607, "step": 2668 }, { "epoch": 0.26, "learning_rate": 0.001728391328155343, "loss": 1.709, "step": 2669 }, { "epoch": 0.26, "learning_rate": 0.0017281725190131584, "loss": 1.5693, "step": 2670 }, { "epoch": 0.26, "learning_rate": 0.0017279536356319754, "loss": 1.7451, "step": 2671 }, { "epoch": 0.26, "learning_rate": 0.0017277346780341092, "loss": 1.7305, "step": 2672 }, { "epoch": 0.26, "learning_rate": 0.0017275156462418836, "loss": 1.6182, "step": 2673 }, { "epoch": 0.26, "learning_rate": 0.0017272965402776288, "loss": 1.8398, "step": 2674 }, { "epoch": 0.26, "learning_rate": 0.0017270773601636839, "loss": 1.7949, "step": 2675 }, { "epoch": 0.26, "learning_rate": 0.001726858105922394, "loss": 1.7178, "step": 2676 }, { "epoch": 0.26, "learning_rate": 0.0017266387775761137, "loss": 1.749, "step": 2677 }, { "epoch": 0.26, "learning_rate": 0.0017264193751472033, "loss": 1.7578, "step": 2678 }, { "epoch": 0.26, "learning_rate": 0.0017261998986580315, "loss": 1.6953, "step": 2679 }, { "epoch": 0.26, "learning_rate": 0.001725980348130975, "loss": 1.7705, "step": 2680 }, { "epoch": 0.26, "learning_rate": 0.001725760723588417, "loss": 1.6729, "step": 2681 }, { "epoch": 0.26, "learning_rate": 0.0017255410250527486, "loss": 1.7725, "step": 2682 }, { "epoch": 0.26, "learning_rate": 0.0017253212525463692, "loss": 1.6621, "step": 2683 }, { "epoch": 0.26, "learning_rate": 0.0017251014060916847, "loss": 1.7725, "step": 2684 }, { "epoch": 0.26, "learning_rate": 0.0017248814857111093, "loss": 1.7617, "step": 2685 }, { "epoch": 0.26, "learning_rate": 0.0017246614914270641, "loss": 1.7432, "step": 2686 }, { "epoch": 0.26, "learning_rate": 0.0017244414232619786, "loss": 1.6514, "step": 2687 }, { "epoch": 0.26, "learning_rate": 0.0017242212812382887, "loss": 1.7295, "step": 2688 }, { "epoch": 0.27, "learning_rate": 0.0017240010653784385, "loss": 1.7324, "step": 2689 }, { "epoch": 0.27, "learning_rate": 0.00172378077570488, "loss": 1.6768, "step": 2690 }, { "epoch": 0.27, "learning_rate": 0.0017235604122400718, "loss": 1.7451, "step": 2691 }, { "epoch": 0.27, "learning_rate": 0.001723339975006481, "loss": 1.7422, "step": 2692 }, { "epoch": 0.27, "learning_rate": 0.0017231194640265812, "loss": 1.7871, "step": 2693 }, { "epoch": 0.27, "learning_rate": 0.0017228988793228544, "loss": 1.7236, "step": 2694 }, { "epoch": 0.27, "learning_rate": 0.0017226782209177895, "loss": 1.7383, "step": 2695 }, { "epoch": 0.27, "learning_rate": 0.0017224574888338835, "loss": 1.7461, "step": 2696 }, { "epoch": 0.27, "learning_rate": 0.0017222366830936405, "loss": 1.6289, "step": 2697 }, { "epoch": 0.27, "learning_rate": 0.001722015803719572, "loss": 1.7676, "step": 2698 }, { "epoch": 0.27, "learning_rate": 0.0017217948507341969, "loss": 1.7773, "step": 2699 }, { "epoch": 0.27, "learning_rate": 0.0017215738241600426, "loss": 1.6279, "step": 2700 }, { "epoch": 0.27, "learning_rate": 0.001721352724019643, "loss": 1.7168, "step": 2701 }, { "epoch": 0.27, "learning_rate": 0.00172113155033554, "loss": 1.8047, "step": 2702 }, { "epoch": 0.27, "learning_rate": 0.0017209103031302826, "loss": 1.7012, "step": 2703 }, { "epoch": 0.27, "learning_rate": 0.0017206889824264273, "loss": 1.793, "step": 2704 }, { "epoch": 0.27, "learning_rate": 0.0017204675882465383, "loss": 1.8008, "step": 2705 }, { "epoch": 0.27, "learning_rate": 0.001720246120613188, "loss": 1.7705, "step": 2706 }, { "epoch": 0.27, "learning_rate": 0.0017200245795489552, "loss": 1.5771, "step": 2707 }, { "epoch": 0.27, "learning_rate": 0.001719802965076426, "loss": 1.7461, "step": 2708 }, { "epoch": 0.27, "learning_rate": 0.001719581277218195, "loss": 1.666, "step": 2709 }, { "epoch": 0.27, "learning_rate": 0.0017193595159968639, "loss": 1.6709, "step": 2710 }, { "epoch": 0.27, "learning_rate": 0.0017191376814350418, "loss": 1.7383, "step": 2711 }, { "epoch": 0.27, "learning_rate": 0.0017189157735553451, "loss": 1.8242, "step": 2712 }, { "epoch": 0.27, "learning_rate": 0.0017186937923803981, "loss": 1.7012, "step": 2713 }, { "epoch": 0.27, "learning_rate": 0.0017184717379328323, "loss": 1.6934, "step": 2714 }, { "epoch": 0.27, "learning_rate": 0.0017182496102352864, "loss": 1.7939, "step": 2715 }, { "epoch": 0.27, "learning_rate": 0.0017180274093104072, "loss": 1.7725, "step": 2716 }, { "epoch": 0.27, "learning_rate": 0.0017178051351808485, "loss": 1.7266, "step": 2717 }, { "epoch": 0.27, "learning_rate": 0.0017175827878692716, "loss": 1.6406, "step": 2718 }, { "epoch": 0.27, "learning_rate": 0.0017173603673983457, "loss": 1.6816, "step": 2719 }, { "epoch": 0.27, "learning_rate": 0.0017171378737907467, "loss": 1.7002, "step": 2720 }, { "epoch": 0.27, "learning_rate": 0.0017169153070691586, "loss": 1.6484, "step": 2721 }, { "epoch": 0.27, "learning_rate": 0.0017166926672562725, "loss": 1.6855, "step": 2722 }, { "epoch": 0.27, "learning_rate": 0.0017164699543747875, "loss": 1.792, "step": 2723 }, { "epoch": 0.27, "learning_rate": 0.0017162471684474092, "loss": 1.666, "step": 2724 }, { "epoch": 0.27, "learning_rate": 0.0017160243094968516, "loss": 1.709, "step": 2725 }, { "epoch": 0.27, "learning_rate": 0.0017158013775458354, "loss": 1.7051, "step": 2726 }, { "epoch": 0.27, "learning_rate": 0.001715578372617089, "loss": 1.7383, "step": 2727 }, { "epoch": 0.27, "learning_rate": 0.0017153552947333486, "loss": 1.6113, "step": 2728 }, { "epoch": 0.27, "learning_rate": 0.0017151321439173577, "loss": 1.7578, "step": 2729 }, { "epoch": 0.27, "learning_rate": 0.0017149089201918665, "loss": 1.6982, "step": 2730 }, { "epoch": 0.27, "learning_rate": 0.0017146856235796338, "loss": 1.583, "step": 2731 }, { "epoch": 0.27, "learning_rate": 0.0017144622541034246, "loss": 1.7676, "step": 2732 }, { "epoch": 0.27, "learning_rate": 0.0017142388117860128, "loss": 1.7695, "step": 2733 }, { "epoch": 0.27, "learning_rate": 0.0017140152966501783, "loss": 1.7441, "step": 2734 }, { "epoch": 0.27, "learning_rate": 0.0017137917087187093, "loss": 1.7822, "step": 2735 }, { "epoch": 0.27, "learning_rate": 0.0017135680480144008, "loss": 1.7432, "step": 2736 }, { "epoch": 0.27, "learning_rate": 0.0017133443145600558, "loss": 1.6992, "step": 2737 }, { "epoch": 0.27, "learning_rate": 0.0017131205083784843, "loss": 1.7354, "step": 2738 }, { "epoch": 0.27, "learning_rate": 0.0017128966294925043, "loss": 1.7207, "step": 2739 }, { "epoch": 0.27, "learning_rate": 0.0017126726779249407, "loss": 1.7139, "step": 2740 }, { "epoch": 0.27, "learning_rate": 0.0017124486536986257, "loss": 1.7227, "step": 2741 }, { "epoch": 0.27, "learning_rate": 0.0017122245568363993, "loss": 1.7842, "step": 2742 }, { "epoch": 0.27, "learning_rate": 0.0017120003873611082, "loss": 1.6494, "step": 2743 }, { "epoch": 0.27, "learning_rate": 0.001711776145295608, "loss": 1.8594, "step": 2744 }, { "epoch": 0.27, "learning_rate": 0.0017115518306627599, "loss": 1.7236, "step": 2745 }, { "epoch": 0.27, "learning_rate": 0.001711327443485434, "loss": 1.7275, "step": 2746 }, { "epoch": 0.27, "learning_rate": 0.0017111029837865064, "loss": 1.8545, "step": 2747 }, { "epoch": 0.27, "learning_rate": 0.001710878451588862, "loss": 1.7246, "step": 2748 }, { "epoch": 0.27, "learning_rate": 0.0017106538469153921, "loss": 1.7158, "step": 2749 }, { "epoch": 0.27, "learning_rate": 0.0017104291697889954, "loss": 1.7598, "step": 2750 }, { "epoch": 0.27, "learning_rate": 0.001710204420232579, "loss": 1.8457, "step": 2751 }, { "epoch": 0.27, "learning_rate": 0.0017099795982690562, "loss": 1.7197, "step": 2752 }, { "epoch": 0.27, "learning_rate": 0.0017097547039213483, "loss": 1.7256, "step": 2753 }, { "epoch": 0.27, "learning_rate": 0.001709529737212384, "loss": 1.8525, "step": 2754 }, { "epoch": 0.27, "learning_rate": 0.0017093046981650987, "loss": 1.6875, "step": 2755 }, { "epoch": 0.27, "learning_rate": 0.001709079586802436, "loss": 1.7295, "step": 2756 }, { "epoch": 0.27, "learning_rate": 0.0017088544031473468, "loss": 1.7705, "step": 2757 }, { "epoch": 0.27, "learning_rate": 0.0017086291472227886, "loss": 1.7617, "step": 2758 }, { "epoch": 0.27, "learning_rate": 0.0017084038190517273, "loss": 1.6074, "step": 2759 }, { "epoch": 0.27, "learning_rate": 0.0017081784186571355, "loss": 1.6104, "step": 2760 }, { "epoch": 0.27, "learning_rate": 0.001707952946061993, "loss": 1.8213, "step": 2761 }, { "epoch": 0.27, "learning_rate": 0.0017077274012892879, "loss": 1.7559, "step": 2762 }, { "epoch": 0.27, "learning_rate": 0.0017075017843620147, "loss": 1.5615, "step": 2763 }, { "epoch": 0.27, "learning_rate": 0.0017072760953031757, "loss": 1.6748, "step": 2764 }, { "epoch": 0.27, "learning_rate": 0.0017070503341357801, "loss": 1.7432, "step": 2765 }, { "epoch": 0.27, "learning_rate": 0.0017068245008828456, "loss": 1.7588, "step": 2766 }, { "epoch": 0.27, "learning_rate": 0.0017065985955673955, "loss": 1.8633, "step": 2767 }, { "epoch": 0.27, "learning_rate": 0.0017063726182124622, "loss": 1.7002, "step": 2768 }, { "epoch": 0.27, "learning_rate": 0.0017061465688410844, "loss": 1.8623, "step": 2769 }, { "epoch": 0.27, "learning_rate": 0.0017059204474763078, "loss": 1.7217, "step": 2770 }, { "epoch": 0.27, "learning_rate": 0.001705694254141187, "loss": 1.7324, "step": 2771 }, { "epoch": 0.27, "learning_rate": 0.0017054679888587828, "loss": 1.7002, "step": 2772 }, { "epoch": 0.27, "learning_rate": 0.001705241651652163, "loss": 1.8467, "step": 2773 }, { "epoch": 0.27, "learning_rate": 0.0017050152425444034, "loss": 1.6895, "step": 2774 }, { "epoch": 0.27, "learning_rate": 0.001704788761558587, "loss": 1.6152, "step": 2775 }, { "epoch": 0.27, "learning_rate": 0.0017045622087178045, "loss": 1.7012, "step": 2776 }, { "epoch": 0.27, "learning_rate": 0.001704335584045153, "loss": 1.9053, "step": 2777 }, { "epoch": 0.27, "learning_rate": 0.0017041088875637375, "loss": 1.6064, "step": 2778 }, { "epoch": 0.27, "learning_rate": 0.0017038821192966707, "loss": 1.542, "step": 2779 }, { "epoch": 0.27, "learning_rate": 0.0017036552792670716, "loss": 1.6348, "step": 2780 }, { "epoch": 0.27, "learning_rate": 0.0017034283674980677, "loss": 1.7764, "step": 2781 }, { "epoch": 0.27, "learning_rate": 0.0017032013840127926, "loss": 1.7061, "step": 2782 }, { "epoch": 0.27, "learning_rate": 0.0017029743288343884, "loss": 1.6924, "step": 2783 }, { "epoch": 0.27, "learning_rate": 0.0017027472019860038, "loss": 1.6016, "step": 2784 }, { "epoch": 0.27, "learning_rate": 0.0017025200034907946, "loss": 1.7178, "step": 2785 }, { "epoch": 0.27, "learning_rate": 0.0017022927333719244, "loss": 1.6807, "step": 2786 }, { "epoch": 0.27, "learning_rate": 0.0017020653916525643, "loss": 1.5166, "step": 2787 }, { "epoch": 0.27, "learning_rate": 0.001701837978355892, "loss": 1.7021, "step": 2788 }, { "epoch": 0.27, "learning_rate": 0.001701610493505093, "loss": 1.7744, "step": 2789 }, { "epoch": 0.28, "learning_rate": 0.00170138293712336, "loss": 1.5664, "step": 2790 }, { "epoch": 0.28, "learning_rate": 0.0017011553092338928, "loss": 1.6426, "step": 2791 }, { "epoch": 0.28, "learning_rate": 0.0017009276098598986, "loss": 1.7344, "step": 2792 }, { "epoch": 0.28, "learning_rate": 0.001700699839024592, "loss": 1.6064, "step": 2793 }, { "epoch": 0.28, "learning_rate": 0.0017004719967511949, "loss": 1.7422, "step": 2794 }, { "epoch": 0.28, "learning_rate": 0.001700244083062936, "loss": 1.7188, "step": 2795 }, { "epoch": 0.28, "learning_rate": 0.0017000160979830522, "loss": 1.6758, "step": 2796 }, { "epoch": 0.28, "learning_rate": 0.0016997880415347867, "loss": 1.7344, "step": 2797 }, { "epoch": 0.28, "learning_rate": 0.0016995599137413905, "loss": 1.8369, "step": 2798 }, { "epoch": 0.28, "learning_rate": 0.0016993317146261223, "loss": 1.6875, "step": 2799 }, { "epoch": 0.28, "learning_rate": 0.0016991034442122466, "loss": 1.8271, "step": 2800 }, { "epoch": 0.28, "learning_rate": 0.001698875102523037, "loss": 1.666, "step": 2801 }, { "epoch": 0.28, "learning_rate": 0.0016986466895817735, "loss": 1.7871, "step": 2802 }, { "epoch": 0.28, "learning_rate": 0.0016984182054117427, "loss": 1.7539, "step": 2803 }, { "epoch": 0.28, "learning_rate": 0.0016981896500362398, "loss": 1.6279, "step": 2804 }, { "epoch": 0.28, "learning_rate": 0.001697961023478566, "loss": 1.7354, "step": 2805 }, { "epoch": 0.28, "learning_rate": 0.0016977323257620308, "loss": 1.6035, "step": 2806 }, { "epoch": 0.28, "learning_rate": 0.0016975035569099504, "loss": 1.6309, "step": 2807 }, { "epoch": 0.28, "learning_rate": 0.001697274716945648, "loss": 1.7822, "step": 2808 }, { "epoch": 0.28, "learning_rate": 0.0016970458058924549, "loss": 1.668, "step": 2809 }, { "epoch": 0.28, "learning_rate": 0.0016968168237737088, "loss": 1.8584, "step": 2810 }, { "epoch": 0.28, "learning_rate": 0.0016965877706127555, "loss": 1.7832, "step": 2811 }, { "epoch": 0.28, "learning_rate": 0.0016963586464329467, "loss": 1.7285, "step": 2812 }, { "epoch": 0.28, "learning_rate": 0.0016961294512576432, "loss": 1.6016, "step": 2813 }, { "epoch": 0.28, "learning_rate": 0.0016959001851102112, "loss": 1.7197, "step": 2814 }, { "epoch": 0.28, "learning_rate": 0.0016956708480140254, "loss": 1.7695, "step": 2815 }, { "epoch": 0.28, "learning_rate": 0.0016954414399924667, "loss": 1.6904, "step": 2816 }, { "epoch": 0.28, "learning_rate": 0.0016952119610689248, "loss": 1.7783, "step": 2817 }, { "epoch": 0.28, "learning_rate": 0.0016949824112667948, "loss": 1.6748, "step": 2818 }, { "epoch": 0.28, "learning_rate": 0.00169475279060948, "loss": 1.6875, "step": 2819 }, { "epoch": 0.28, "learning_rate": 0.0016945230991203912, "loss": 1.7412, "step": 2820 }, { "epoch": 0.28, "learning_rate": 0.0016942933368229455, "loss": 1.6777, "step": 2821 }, { "epoch": 0.28, "learning_rate": 0.0016940635037405686, "loss": 1.6846, "step": 2822 }, { "epoch": 0.28, "learning_rate": 0.0016938335998966916, "loss": 1.7227, "step": 2823 }, { "epoch": 0.28, "learning_rate": 0.0016936036253147539, "loss": 1.666, "step": 2824 }, { "epoch": 0.28, "learning_rate": 0.0016933735800182027, "loss": 1.623, "step": 2825 }, { "epoch": 0.28, "learning_rate": 0.0016931434640304905, "loss": 1.6953, "step": 2826 }, { "epoch": 0.28, "learning_rate": 0.0016929132773750798, "loss": 1.6475, "step": 2827 }, { "epoch": 0.28, "learning_rate": 0.0016926830200754371, "loss": 1.6465, "step": 2828 }, { "epoch": 0.28, "learning_rate": 0.0016924526921550387, "loss": 1.8926, "step": 2829 }, { "epoch": 0.28, "learning_rate": 0.0016922222936373665, "loss": 1.5967, "step": 2830 }, { "epoch": 0.28, "learning_rate": 0.001691991824545911, "loss": 1.7158, "step": 2831 }, { "epoch": 0.28, "learning_rate": 0.0016917612849041684, "loss": 1.6592, "step": 2832 }, { "epoch": 0.28, "learning_rate": 0.0016915306747356431, "loss": 1.7354, "step": 2833 }, { "epoch": 0.28, "learning_rate": 0.0016912999940638463, "loss": 1.5781, "step": 2834 }, { "epoch": 0.28, "learning_rate": 0.0016910692429122963, "loss": 1.7285, "step": 2835 }, { "epoch": 0.28, "learning_rate": 0.0016908384213045187, "loss": 1.8115, "step": 2836 }, { "epoch": 0.28, "learning_rate": 0.0016906075292640469, "loss": 1.6514, "step": 2837 }, { "epoch": 0.28, "learning_rate": 0.0016903765668144204, "loss": 1.5996, "step": 2838 }, { "epoch": 0.28, "learning_rate": 0.0016901455339791865, "loss": 1.6875, "step": 2839 }, { "epoch": 0.28, "learning_rate": 0.0016899144307818998, "loss": 1.75, "step": 2840 }, { "epoch": 0.28, "learning_rate": 0.0016896832572461215, "loss": 1.8057, "step": 2841 }, { "epoch": 0.28, "learning_rate": 0.0016894520133954205, "loss": 1.6514, "step": 2842 }, { "epoch": 0.28, "learning_rate": 0.0016892206992533728, "loss": 1.6934, "step": 2843 }, { "epoch": 0.28, "learning_rate": 0.0016889893148435613, "loss": 1.7461, "step": 2844 }, { "epoch": 0.28, "learning_rate": 0.0016887578601895764, "loss": 1.7812, "step": 2845 }, { "epoch": 0.28, "learning_rate": 0.0016885263353150148, "loss": 1.8896, "step": 2846 }, { "epoch": 0.28, "learning_rate": 0.001688294740243482, "loss": 1.752, "step": 2847 }, { "epoch": 0.28, "learning_rate": 0.0016880630749985892, "loss": 1.6299, "step": 2848 }, { "epoch": 0.28, "learning_rate": 0.001687831339603955, "loss": 1.6367, "step": 2849 }, { "epoch": 0.28, "learning_rate": 0.0016875995340832062, "loss": 1.7178, "step": 2850 }, { "epoch": 0.28, "learning_rate": 0.001687367658459975, "loss": 1.6543, "step": 2851 }, { "epoch": 0.28, "learning_rate": 0.0016871357127579026, "loss": 1.7285, "step": 2852 }, { "epoch": 0.28, "learning_rate": 0.0016869036970006356, "loss": 1.6182, "step": 2853 }, { "epoch": 0.28, "learning_rate": 0.001686671611211829, "loss": 1.7656, "step": 2854 }, { "epoch": 0.28, "learning_rate": 0.0016864394554151443, "loss": 1.6289, "step": 2855 }, { "epoch": 0.28, "learning_rate": 0.001686207229634251, "loss": 1.7373, "step": 2856 }, { "epoch": 0.28, "learning_rate": 0.0016859749338928242, "loss": 1.6582, "step": 2857 }, { "epoch": 0.28, "learning_rate": 0.0016857425682145477, "loss": 1.8604, "step": 2858 }, { "epoch": 0.28, "learning_rate": 0.0016855101326231118, "loss": 1.6406, "step": 2859 }, { "epoch": 0.28, "learning_rate": 0.0016852776271422132, "loss": 1.7197, "step": 2860 }, { "epoch": 0.28, "learning_rate": 0.0016850450517955568, "loss": 1.707, "step": 2861 }, { "epoch": 0.28, "learning_rate": 0.0016848124066068547, "loss": 1.6377, "step": 2862 }, { "epoch": 0.28, "learning_rate": 0.0016845796915998248, "loss": 1.6035, "step": 2863 }, { "epoch": 0.28, "learning_rate": 0.0016843469067981936, "loss": 1.7764, "step": 2864 }, { "epoch": 0.28, "learning_rate": 0.0016841140522256938, "loss": 1.7236, "step": 2865 }, { "epoch": 0.28, "learning_rate": 0.0016838811279060656, "loss": 1.6699, "step": 2866 }, { "epoch": 0.28, "learning_rate": 0.0016836481338630565, "loss": 1.6904, "step": 2867 }, { "epoch": 0.28, "learning_rate": 0.0016834150701204203, "loss": 1.7314, "step": 2868 }, { "epoch": 0.28, "learning_rate": 0.0016831819367019182, "loss": 1.6885, "step": 2869 }, { "epoch": 0.28, "learning_rate": 0.0016829487336313198, "loss": 1.748, "step": 2870 }, { "epoch": 0.28, "learning_rate": 0.0016827154609323999, "loss": 1.8057, "step": 2871 }, { "epoch": 0.28, "learning_rate": 0.0016824821186289414, "loss": 1.6689, "step": 2872 }, { "epoch": 0.28, "learning_rate": 0.001682248706744734, "loss": 1.7949, "step": 2873 }, { "epoch": 0.28, "learning_rate": 0.0016820152253035753, "loss": 1.7998, "step": 2874 }, { "epoch": 0.28, "learning_rate": 0.0016817816743292685, "loss": 1.7402, "step": 2875 }, { "epoch": 0.28, "learning_rate": 0.0016815480538456247, "loss": 1.6572, "step": 2876 }, { "epoch": 0.28, "learning_rate": 0.0016813143638764628, "loss": 1.7041, "step": 2877 }, { "epoch": 0.28, "learning_rate": 0.0016810806044456077, "loss": 1.585, "step": 2878 }, { "epoch": 0.28, "learning_rate": 0.0016808467755768917, "loss": 1.7676, "step": 2879 }, { "epoch": 0.28, "learning_rate": 0.0016806128772941542, "loss": 1.7607, "step": 2880 }, { "epoch": 0.28, "learning_rate": 0.0016803789096212416, "loss": 1.751, "step": 2881 }, { "epoch": 0.28, "learning_rate": 0.001680144872582008, "loss": 1.623, "step": 2882 }, { "epoch": 0.28, "learning_rate": 0.0016799107662003134, "loss": 1.7021, "step": 2883 }, { "epoch": 0.28, "learning_rate": 0.001679676590500026, "loss": 1.6279, "step": 2884 }, { "epoch": 0.28, "learning_rate": 0.0016794423455050204, "loss": 1.8057, "step": 2885 }, { "epoch": 0.28, "learning_rate": 0.0016792080312391785, "loss": 1.7666, "step": 2886 }, { "epoch": 0.28, "learning_rate": 0.001678973647726389, "loss": 1.7754, "step": 2887 }, { "epoch": 0.28, "learning_rate": 0.0016787391949905483, "loss": 1.7012, "step": 2888 }, { "epoch": 0.28, "learning_rate": 0.0016785046730555596, "loss": 1.7832, "step": 2889 }, { "epoch": 0.28, "learning_rate": 0.0016782700819453322, "loss": 1.7236, "step": 2890 }, { "epoch": 0.28, "learning_rate": 0.0016780354216837838, "loss": 1.5625, "step": 2891 }, { "epoch": 0.29, "learning_rate": 0.0016778006922948383, "loss": 1.7207, "step": 2892 }, { "epoch": 0.29, "learning_rate": 0.0016775658938024277, "loss": 1.7578, "step": 2893 }, { "epoch": 0.29, "learning_rate": 0.0016773310262304891, "loss": 1.6191, "step": 2894 }, { "epoch": 0.29, "learning_rate": 0.0016770960896029686, "loss": 1.6846, "step": 2895 }, { "epoch": 0.29, "learning_rate": 0.0016768610839438185, "loss": 1.7285, "step": 2896 }, { "epoch": 0.29, "learning_rate": 0.001676626009276998, "loss": 1.7646, "step": 2897 }, { "epoch": 0.29, "learning_rate": 0.001676390865626474, "loss": 1.7861, "step": 2898 }, { "epoch": 0.29, "learning_rate": 0.0016761556530162194, "loss": 1.6641, "step": 2899 }, { "epoch": 0.29, "learning_rate": 0.001675920371470215, "loss": 1.6836, "step": 2900 }, { "epoch": 0.29, "learning_rate": 0.0016756850210124484, "loss": 1.6543, "step": 2901 }, { "epoch": 0.29, "learning_rate": 0.001675449601666914, "loss": 1.708, "step": 2902 }, { "epoch": 0.29, "learning_rate": 0.0016752141134576133, "loss": 1.6924, "step": 2903 }, { "epoch": 0.29, "learning_rate": 0.0016749785564085554, "loss": 1.6172, "step": 2904 }, { "epoch": 0.29, "learning_rate": 0.0016747429305437554, "loss": 1.7559, "step": 2905 }, { "epoch": 0.29, "learning_rate": 0.0016745072358872362, "loss": 1.7148, "step": 2906 }, { "epoch": 0.29, "learning_rate": 0.0016742714724630274, "loss": 1.6455, "step": 2907 }, { "epoch": 0.29, "learning_rate": 0.0016740356402951658, "loss": 1.6934, "step": 2908 }, { "epoch": 0.29, "learning_rate": 0.0016737997394076945, "loss": 1.7363, "step": 2909 }, { "epoch": 0.29, "learning_rate": 0.001673563769824665, "loss": 1.7891, "step": 2910 }, { "epoch": 0.29, "learning_rate": 0.0016733277315701346, "loss": 1.6709, "step": 2911 }, { "epoch": 0.29, "learning_rate": 0.0016730916246681677, "loss": 1.7793, "step": 2912 }, { "epoch": 0.29, "learning_rate": 0.0016728554491428364, "loss": 1.7793, "step": 2913 }, { "epoch": 0.29, "learning_rate": 0.0016726192050182194, "loss": 1.8008, "step": 2914 }, { "epoch": 0.29, "learning_rate": 0.0016723828923184021, "loss": 1.748, "step": 2915 }, { "epoch": 0.29, "learning_rate": 0.0016721465110674774, "loss": 1.7441, "step": 2916 }, { "epoch": 0.29, "learning_rate": 0.0016719100612895448, "loss": 1.7002, "step": 2917 }, { "epoch": 0.29, "learning_rate": 0.0016716735430087112, "loss": 1.6631, "step": 2918 }, { "epoch": 0.29, "learning_rate": 0.0016714369562490896, "loss": 1.7168, "step": 2919 }, { "epoch": 0.29, "learning_rate": 0.0016712003010348013, "loss": 1.6836, "step": 2920 }, { "epoch": 0.29, "learning_rate": 0.0016709635773899736, "loss": 1.7168, "step": 2921 }, { "epoch": 0.29, "learning_rate": 0.0016707267853387413, "loss": 1.6094, "step": 2922 }, { "epoch": 0.29, "learning_rate": 0.0016704899249052455, "loss": 1.7656, "step": 2923 }, { "epoch": 0.29, "learning_rate": 0.0016702529961136348, "loss": 1.6201, "step": 2924 }, { "epoch": 0.29, "learning_rate": 0.001670015998988065, "loss": 1.7686, "step": 2925 }, { "epoch": 0.29, "learning_rate": 0.0016697789335526983, "loss": 1.7227, "step": 2926 }, { "epoch": 0.29, "learning_rate": 0.0016695417998317042, "loss": 1.6914, "step": 2927 }, { "epoch": 0.29, "learning_rate": 0.001669304597849259, "loss": 1.8613, "step": 2928 }, { "epoch": 0.29, "learning_rate": 0.0016690673276295463, "loss": 1.6777, "step": 2929 }, { "epoch": 0.29, "learning_rate": 0.0016688299891967557, "loss": 1.7402, "step": 2930 }, { "epoch": 0.29, "learning_rate": 0.0016685925825750854, "loss": 1.6719, "step": 2931 }, { "epoch": 0.29, "learning_rate": 0.001668355107788739, "loss": 1.7295, "step": 2932 }, { "epoch": 0.29, "learning_rate": 0.0016681175648619276, "loss": 1.6816, "step": 2933 }, { "epoch": 0.29, "learning_rate": 0.0016678799538188695, "loss": 1.6455, "step": 2934 }, { "epoch": 0.29, "learning_rate": 0.0016676422746837898, "loss": 1.7412, "step": 2935 }, { "epoch": 0.29, "learning_rate": 0.0016674045274809202, "loss": 1.5947, "step": 2936 }, { "epoch": 0.29, "learning_rate": 0.0016671667122344998, "loss": 1.7012, "step": 2937 }, { "epoch": 0.29, "learning_rate": 0.0016669288289687745, "loss": 1.7979, "step": 2938 }, { "epoch": 0.29, "learning_rate": 0.001666690877707997, "loss": 1.7354, "step": 2939 }, { "epoch": 0.29, "learning_rate": 0.001666452858476427, "loss": 1.7344, "step": 2940 }, { "epoch": 0.29, "learning_rate": 0.001666214771298331, "loss": 1.6973, "step": 2941 }, { "epoch": 0.29, "learning_rate": 0.001665976616197983, "loss": 1.7285, "step": 2942 }, { "epoch": 0.29, "learning_rate": 0.0016657383931996633, "loss": 1.6582, "step": 2943 }, { "epoch": 0.29, "learning_rate": 0.001665500102327659, "loss": 1.7246, "step": 2944 }, { "epoch": 0.29, "learning_rate": 0.0016652617436062652, "loss": 1.6084, "step": 2945 }, { "epoch": 0.29, "learning_rate": 0.0016650233170597826, "loss": 1.6758, "step": 2946 }, { "epoch": 0.29, "learning_rate": 0.0016647848227125195, "loss": 1.6367, "step": 2947 }, { "epoch": 0.29, "learning_rate": 0.0016645462605887906, "loss": 1.7598, "step": 2948 }, { "epoch": 0.29, "learning_rate": 0.0016643076307129188, "loss": 1.7451, "step": 2949 }, { "epoch": 0.29, "learning_rate": 0.0016640689331092323, "loss": 1.7832, "step": 2950 }, { "epoch": 0.29, "learning_rate": 0.001663830167802067, "loss": 1.6182, "step": 2951 }, { "epoch": 0.29, "learning_rate": 0.0016635913348157661, "loss": 1.6357, "step": 2952 }, { "epoch": 0.29, "learning_rate": 0.001663352434174678, "loss": 1.7725, "step": 2953 }, { "epoch": 0.29, "learning_rate": 0.001663113465903161, "loss": 1.6475, "step": 2954 }, { "epoch": 0.29, "learning_rate": 0.0016628744300255774, "loss": 1.7148, "step": 2955 }, { "epoch": 0.29, "learning_rate": 0.0016626353265662974, "loss": 1.7578, "step": 2956 }, { "epoch": 0.29, "learning_rate": 0.0016623961555496988, "loss": 1.6211, "step": 2957 }, { "epoch": 0.29, "learning_rate": 0.0016621569170001647, "loss": 1.8359, "step": 2958 }, { "epoch": 0.29, "learning_rate": 0.001661917610942087, "loss": 1.6094, "step": 2959 }, { "epoch": 0.29, "learning_rate": 0.0016616782373998636, "loss": 1.8252, "step": 2960 }, { "epoch": 0.29, "learning_rate": 0.0016614387963978987, "loss": 1.6768, "step": 2961 }, { "epoch": 0.29, "learning_rate": 0.001661199287960604, "loss": 1.7568, "step": 2962 }, { "epoch": 0.29, "learning_rate": 0.0016609597121123982, "loss": 1.6875, "step": 2963 }, { "epoch": 0.29, "learning_rate": 0.0016607200688777066, "loss": 1.7168, "step": 2964 }, { "epoch": 0.29, "learning_rate": 0.001660480358280961, "loss": 1.7178, "step": 2965 }, { "epoch": 0.29, "learning_rate": 0.001660240580346601, "loss": 1.6621, "step": 2966 }, { "epoch": 0.29, "learning_rate": 0.0016600007350990726, "loss": 1.7861, "step": 2967 }, { "epoch": 0.29, "learning_rate": 0.001659760822562828, "loss": 1.7471, "step": 2968 }, { "epoch": 0.29, "learning_rate": 0.0016595208427623276, "loss": 1.7383, "step": 2969 }, { "epoch": 0.29, "learning_rate": 0.0016592807957220375, "loss": 1.6299, "step": 2970 }, { "epoch": 0.29, "learning_rate": 0.0016590406814664313, "loss": 1.7148, "step": 2971 }, { "epoch": 0.29, "learning_rate": 0.0016588005000199894, "loss": 1.6279, "step": 2972 }, { "epoch": 0.29, "learning_rate": 0.0016585602514071983, "loss": 1.7217, "step": 2973 }, { "epoch": 0.29, "learning_rate": 0.0016583199356525525, "loss": 1.6758, "step": 2974 }, { "epoch": 0.29, "learning_rate": 0.0016580795527805523, "loss": 1.6865, "step": 2975 }, { "epoch": 0.29, "learning_rate": 0.0016578391028157057, "loss": 1.6865, "step": 2976 }, { "epoch": 0.29, "learning_rate": 0.0016575985857825274, "loss": 1.7871, "step": 2977 }, { "epoch": 0.29, "learning_rate": 0.0016573580017055383, "loss": 1.6963, "step": 2978 }, { "epoch": 0.29, "learning_rate": 0.0016571173506092665, "loss": 1.5713, "step": 2979 }, { "epoch": 0.29, "learning_rate": 0.0016568766325182475, "loss": 1.7529, "step": 2980 }, { "epoch": 0.29, "learning_rate": 0.0016566358474570224, "loss": 1.7627, "step": 2981 }, { "epoch": 0.29, "learning_rate": 0.0016563949954501403, "loss": 1.7227, "step": 2982 }, { "epoch": 0.29, "learning_rate": 0.0016561540765221566, "loss": 1.5391, "step": 2983 }, { "epoch": 0.29, "learning_rate": 0.0016559130906976334, "loss": 1.6582, "step": 2984 }, { "epoch": 0.29, "learning_rate": 0.00165567203800114, "loss": 1.8252, "step": 2985 }, { "epoch": 0.29, "learning_rate": 0.001655430918457252, "loss": 1.7246, "step": 2986 }, { "epoch": 0.29, "learning_rate": 0.0016551897320905527, "loss": 1.8193, "step": 2987 }, { "epoch": 0.29, "learning_rate": 0.001654948478925631, "loss": 1.7588, "step": 2988 }, { "epoch": 0.29, "learning_rate": 0.0016547071589870838, "loss": 1.7236, "step": 2989 }, { "epoch": 0.29, "learning_rate": 0.0016544657722995142, "loss": 1.7324, "step": 2990 }, { "epoch": 0.29, "learning_rate": 0.001654224318887532, "loss": 1.7344, "step": 2991 }, { "epoch": 0.29, "learning_rate": 0.001653982798775754, "loss": 1.6201, "step": 2992 }, { "epoch": 0.3, "learning_rate": 0.0016537412119888035, "loss": 1.7197, "step": 2993 }, { "epoch": 0.3, "learning_rate": 0.0016534995585513114, "loss": 1.7988, "step": 2994 }, { "epoch": 0.3, "learning_rate": 0.0016532578384879145, "loss": 1.6768, "step": 2995 }, { "epoch": 0.3, "learning_rate": 0.0016530160518232568, "loss": 1.5967, "step": 2996 }, { "epoch": 0.3, "learning_rate": 0.0016527741985819895, "loss": 1.6152, "step": 2997 }, { "epoch": 0.3, "learning_rate": 0.0016525322787887696, "loss": 1.8828, "step": 2998 }, { "epoch": 0.3, "learning_rate": 0.0016522902924682615, "loss": 1.6787, "step": 2999 }, { "epoch": 0.3, "learning_rate": 0.0016520482396451363, "loss": 1.6953, "step": 3000 }, { "epoch": 0.3, "learning_rate": 0.0016518061203440722, "loss": 1.6074, "step": 3001 }, { "epoch": 0.3, "learning_rate": 0.0016515639345897537, "loss": 1.7354, "step": 3002 }, { "epoch": 0.3, "learning_rate": 0.0016513216824068722, "loss": 1.6953, "step": 3003 }, { "epoch": 0.3, "learning_rate": 0.001651079363820126, "loss": 1.6738, "step": 3004 }, { "epoch": 0.3, "learning_rate": 0.0016508369788542196, "loss": 1.7236, "step": 3005 }, { "epoch": 0.3, "learning_rate": 0.0016505945275338654, "loss": 1.7539, "step": 3006 }, { "epoch": 0.3, "learning_rate": 0.0016503520098837816, "loss": 1.7725, "step": 3007 }, { "epoch": 0.3, "learning_rate": 0.0016501094259286937, "loss": 1.6514, "step": 3008 }, { "epoch": 0.3, "learning_rate": 0.0016498667756933334, "loss": 1.6826, "step": 3009 }, { "epoch": 0.3, "learning_rate": 0.0016496240592024398, "loss": 1.584, "step": 3010 }, { "epoch": 0.3, "learning_rate": 0.0016493812764807583, "loss": 1.7217, "step": 3011 }, { "epoch": 0.3, "learning_rate": 0.0016491384275530412, "loss": 1.6221, "step": 3012 }, { "epoch": 0.3, "learning_rate": 0.0016488955124440475, "loss": 1.6367, "step": 3013 }, { "epoch": 0.3, "learning_rate": 0.0016486525311785432, "loss": 1.6709, "step": 3014 }, { "epoch": 0.3, "learning_rate": 0.0016484094837813006, "loss": 1.667, "step": 3015 }, { "epoch": 0.3, "learning_rate": 0.001648166370277099, "loss": 1.6367, "step": 3016 }, { "epoch": 0.3, "learning_rate": 0.0016479231906907249, "loss": 1.7471, "step": 3017 }, { "epoch": 0.3, "learning_rate": 0.0016476799450469706, "loss": 1.7012, "step": 3018 }, { "epoch": 0.3, "learning_rate": 0.0016474366333706354, "loss": 1.6924, "step": 3019 }, { "epoch": 0.3, "learning_rate": 0.001647193255686526, "loss": 1.6309, "step": 3020 }, { "epoch": 0.3, "learning_rate": 0.0016469498120194553, "loss": 1.7344, "step": 3021 }, { "epoch": 0.3, "learning_rate": 0.0016467063023942426, "loss": 1.6855, "step": 3022 }, { "epoch": 0.3, "learning_rate": 0.0016464627268357149, "loss": 1.7275, "step": 3023 }, { "epoch": 0.3, "learning_rate": 0.0016462190853687047, "loss": 1.7109, "step": 3024 }, { "epoch": 0.3, "learning_rate": 0.0016459753780180524, "loss": 1.8057, "step": 3025 }, { "epoch": 0.3, "learning_rate": 0.0016457316048086041, "loss": 1.7256, "step": 3026 }, { "epoch": 0.3, "learning_rate": 0.0016454877657652136, "loss": 1.7939, "step": 3027 }, { "epoch": 0.3, "learning_rate": 0.0016452438609127404, "loss": 1.7109, "step": 3028 }, { "epoch": 0.3, "learning_rate": 0.0016449998902760515, "loss": 1.5889, "step": 3029 }, { "epoch": 0.3, "learning_rate": 0.0016447558538800202, "loss": 1.7861, "step": 3030 }, { "epoch": 0.3, "learning_rate": 0.0016445117517495268, "loss": 1.7793, "step": 3031 }, { "epoch": 0.3, "learning_rate": 0.001644267583909458, "loss": 1.5518, "step": 3032 }, { "epoch": 0.3, "learning_rate": 0.0016440233503847068, "loss": 1.6523, "step": 3033 }, { "epoch": 0.3, "learning_rate": 0.0016437790512001745, "loss": 1.748, "step": 3034 }, { "epoch": 0.3, "learning_rate": 0.0016435346863807672, "loss": 1.6572, "step": 3035 }, { "epoch": 0.3, "learning_rate": 0.0016432902559513987, "loss": 1.7305, "step": 3036 }, { "epoch": 0.3, "learning_rate": 0.0016430457599369893, "loss": 1.6943, "step": 3037 }, { "epoch": 0.3, "learning_rate": 0.001642801198362466, "loss": 1.6689, "step": 3038 }, { "epoch": 0.3, "learning_rate": 0.0016425565712527624, "loss": 1.7266, "step": 3039 }, { "epoch": 0.3, "learning_rate": 0.0016423118786328189, "loss": 1.6309, "step": 3040 }, { "epoch": 0.3, "learning_rate": 0.0016420671205275826, "loss": 1.8271, "step": 3041 }, { "epoch": 0.3, "learning_rate": 0.0016418222969620069, "loss": 1.7754, "step": 3042 }, { "epoch": 0.3, "learning_rate": 0.0016415774079610525, "loss": 1.6465, "step": 3043 }, { "epoch": 0.3, "learning_rate": 0.0016413324535496864, "loss": 1.7256, "step": 3044 }, { "epoch": 0.3, "learning_rate": 0.0016410874337528822, "loss": 1.7646, "step": 3045 }, { "epoch": 0.3, "learning_rate": 0.0016408423485956205, "loss": 1.7959, "step": 3046 }, { "epoch": 0.3, "learning_rate": 0.001640597198102888, "loss": 1.6514, "step": 3047 }, { "epoch": 0.3, "learning_rate": 0.0016403519822996784, "loss": 1.6748, "step": 3048 }, { "epoch": 0.3, "learning_rate": 0.0016401067012109922, "loss": 1.6514, "step": 3049 }, { "epoch": 0.3, "learning_rate": 0.0016398613548618366, "loss": 1.6768, "step": 3050 }, { "epoch": 0.3, "learning_rate": 0.0016396159432772249, "loss": 1.7256, "step": 3051 }, { "epoch": 0.3, "learning_rate": 0.0016393704664821778, "loss": 1.623, "step": 3052 }, { "epoch": 0.3, "learning_rate": 0.0016391249245017222, "loss": 1.6709, "step": 3053 }, { "epoch": 0.3, "learning_rate": 0.001638879317360891, "loss": 1.5645, "step": 3054 }, { "epoch": 0.3, "learning_rate": 0.0016386336450847257, "loss": 1.8076, "step": 3055 }, { "epoch": 0.3, "learning_rate": 0.0016383879076982725, "loss": 1.7197, "step": 3056 }, { "epoch": 0.3, "learning_rate": 0.0016381421052265848, "loss": 1.6865, "step": 3057 }, { "epoch": 0.3, "learning_rate": 0.001637896237694723, "loss": 1.7842, "step": 3058 }, { "epoch": 0.3, "learning_rate": 0.001637650305127754, "loss": 1.626, "step": 3059 }, { "epoch": 0.3, "learning_rate": 0.001637404307550751, "loss": 1.8311, "step": 3060 }, { "epoch": 0.3, "learning_rate": 0.0016371582449887941, "loss": 1.6377, "step": 3061 }, { "epoch": 0.3, "learning_rate": 0.00163691211746697, "loss": 1.6836, "step": 3062 }, { "epoch": 0.3, "learning_rate": 0.0016366659250103726, "loss": 1.5615, "step": 3063 }, { "epoch": 0.3, "learning_rate": 0.001636419667644101, "loss": 1.6074, "step": 3064 }, { "epoch": 0.3, "learning_rate": 0.0016361733453932618, "loss": 1.8184, "step": 3065 }, { "epoch": 0.3, "learning_rate": 0.0016359269582829688, "loss": 1.7402, "step": 3066 }, { "epoch": 0.3, "learning_rate": 0.0016356805063383413, "loss": 1.6895, "step": 3067 }, { "epoch": 0.3, "learning_rate": 0.0016354339895845059, "loss": 1.6738, "step": 3068 }, { "epoch": 0.3, "learning_rate": 0.0016351874080465954, "loss": 1.6338, "step": 3069 }, { "epoch": 0.3, "learning_rate": 0.0016349407617497493, "loss": 1.7598, "step": 3070 }, { "epoch": 0.3, "learning_rate": 0.0016346940507191144, "loss": 1.6484, "step": 3071 }, { "epoch": 0.3, "learning_rate": 0.0016344472749798427, "loss": 1.5986, "step": 3072 }, { "epoch": 0.3, "learning_rate": 0.0016342004345570942, "loss": 1.6455, "step": 3073 }, { "epoch": 0.3, "learning_rate": 0.0016339535294760347, "loss": 1.79, "step": 3074 }, { "epoch": 0.3, "learning_rate": 0.0016337065597618369, "loss": 1.7363, "step": 3075 }, { "epoch": 0.3, "learning_rate": 0.0016334595254396795, "loss": 1.7031, "step": 3076 }, { "epoch": 0.3, "learning_rate": 0.001633212426534749, "loss": 1.7646, "step": 3077 }, { "epoch": 0.3, "learning_rate": 0.0016329652630722373, "loss": 1.6523, "step": 3078 }, { "epoch": 0.3, "learning_rate": 0.0016327180350773432, "loss": 1.6992, "step": 3079 }, { "epoch": 0.3, "learning_rate": 0.0016324707425752726, "loss": 1.834, "step": 3080 }, { "epoch": 0.3, "learning_rate": 0.0016322233855912375, "loss": 1.7842, "step": 3081 }, { "epoch": 0.3, "learning_rate": 0.0016319759641504562, "loss": 1.5684, "step": 3082 }, { "epoch": 0.3, "learning_rate": 0.0016317284782781543, "loss": 1.5596, "step": 3083 }, { "epoch": 0.3, "learning_rate": 0.0016314809279995633, "loss": 1.7881, "step": 3084 }, { "epoch": 0.3, "learning_rate": 0.0016312333133399219, "loss": 1.6309, "step": 3085 }, { "epoch": 0.3, "learning_rate": 0.001630985634324475, "loss": 1.7705, "step": 3086 }, { "epoch": 0.3, "learning_rate": 0.0016307378909784739, "loss": 1.8467, "step": 3087 }, { "epoch": 0.3, "learning_rate": 0.0016304900833271766, "loss": 1.8086, "step": 3088 }, { "epoch": 0.3, "learning_rate": 0.001630242211395848, "loss": 1.6064, "step": 3089 }, { "epoch": 0.3, "learning_rate": 0.001629994275209759, "loss": 1.7334, "step": 3090 }, { "epoch": 0.3, "learning_rate": 0.0016297462747941875, "loss": 1.6172, "step": 3091 }, { "epoch": 0.3, "learning_rate": 0.0016294982101744174, "loss": 1.7861, "step": 3092 }, { "epoch": 0.3, "learning_rate": 0.0016292500813757398, "loss": 1.79, "step": 3093 }, { "epoch": 0.31, "learning_rate": 0.0016290018884234524, "loss": 1.5645, "step": 3094 }, { "epoch": 0.31, "learning_rate": 0.0016287536313428581, "loss": 1.6377, "step": 3095 }, { "epoch": 0.31, "learning_rate": 0.0016285053101592686, "loss": 1.7861, "step": 3096 }, { "epoch": 0.31, "learning_rate": 0.0016282569248979998, "loss": 1.7139, "step": 3097 }, { "epoch": 0.31, "learning_rate": 0.0016280084755843755, "loss": 1.8271, "step": 3098 }, { "epoch": 0.31, "learning_rate": 0.001627759962243726, "loss": 1.748, "step": 3099 }, { "epoch": 0.31, "learning_rate": 0.0016275113849013876, "loss": 1.7393, "step": 3100 }, { "epoch": 0.31, "learning_rate": 0.0016272627435827034, "loss": 1.7451, "step": 3101 }, { "epoch": 0.31, "learning_rate": 0.0016270140383130229, "loss": 1.7139, "step": 3102 }, { "epoch": 0.31, "learning_rate": 0.001626765269117703, "loss": 1.6602, "step": 3103 }, { "epoch": 0.31, "learning_rate": 0.0016265164360221053, "loss": 1.6787, "step": 3104 }, { "epoch": 0.31, "learning_rate": 0.0016262675390515994, "loss": 1.7129, "step": 3105 }, { "epoch": 0.31, "learning_rate": 0.0016260185782315606, "loss": 1.6699, "step": 3106 }, { "epoch": 0.31, "learning_rate": 0.0016257695535873715, "loss": 1.7607, "step": 3107 }, { "epoch": 0.31, "learning_rate": 0.0016255204651444211, "loss": 1.5215, "step": 3108 }, { "epoch": 0.31, "learning_rate": 0.0016252713129281035, "loss": 1.7354, "step": 3109 }, { "epoch": 0.31, "learning_rate": 0.0016250220969638214, "loss": 1.6855, "step": 3110 }, { "epoch": 0.31, "learning_rate": 0.0016247728172769827, "loss": 1.7314, "step": 3111 }, { "epoch": 0.31, "learning_rate": 0.0016245234738930022, "loss": 1.7656, "step": 3112 }, { "epoch": 0.31, "learning_rate": 0.0016242740668373004, "loss": 1.5957, "step": 3113 }, { "epoch": 0.31, "learning_rate": 0.0016240245961353058, "loss": 1.8301, "step": 3114 }, { "epoch": 0.31, "learning_rate": 0.0016237750618124521, "loss": 1.7715, "step": 3115 }, { "epoch": 0.31, "learning_rate": 0.00162352546389418, "loss": 1.7119, "step": 3116 }, { "epoch": 0.31, "learning_rate": 0.0016232758024059368, "loss": 1.667, "step": 3117 }, { "epoch": 0.31, "learning_rate": 0.001623026077373176, "loss": 1.8242, "step": 3118 }, { "epoch": 0.31, "learning_rate": 0.0016227762888213575, "loss": 1.7568, "step": 3119 }, { "epoch": 0.31, "learning_rate": 0.0016225264367759485, "loss": 1.6719, "step": 3120 }, { "epoch": 0.31, "learning_rate": 0.0016222765212624212, "loss": 1.6719, "step": 3121 }, { "epoch": 0.31, "learning_rate": 0.0016220265423062556, "loss": 1.7256, "step": 3122 }, { "epoch": 0.31, "learning_rate": 0.0016217764999329378, "loss": 1.5557, "step": 3123 }, { "epoch": 0.31, "learning_rate": 0.0016215263941679599, "loss": 1.6963, "step": 3124 }, { "epoch": 0.31, "learning_rate": 0.0016212762250368206, "loss": 1.7031, "step": 3125 }, { "epoch": 0.31, "learning_rate": 0.0016210259925650262, "loss": 1.6914, "step": 3126 }, { "epoch": 0.31, "learning_rate": 0.0016207756967780873, "loss": 1.7314, "step": 3127 }, { "epoch": 0.31, "learning_rate": 0.001620525337701523, "loss": 1.7861, "step": 3128 }, { "epoch": 0.31, "learning_rate": 0.001620274915360858, "loss": 1.6963, "step": 3129 }, { "epoch": 0.31, "learning_rate": 0.0016200244297816231, "loss": 1.5146, "step": 3130 }, { "epoch": 0.31, "learning_rate": 0.0016197738809893563, "loss": 1.7051, "step": 3131 }, { "epoch": 0.31, "learning_rate": 0.0016195232690096014, "loss": 1.6738, "step": 3132 }, { "epoch": 0.31, "learning_rate": 0.0016192725938679092, "loss": 1.6182, "step": 3133 }, { "epoch": 0.31, "learning_rate": 0.001619021855589836, "loss": 1.6445, "step": 3134 }, { "epoch": 0.31, "learning_rate": 0.0016187710542009462, "loss": 1.8096, "step": 3135 }, { "epoch": 0.31, "learning_rate": 0.0016185201897268087, "loss": 1.7188, "step": 3136 }, { "epoch": 0.31, "learning_rate": 0.0016182692621930005, "loss": 1.7803, "step": 3137 }, { "epoch": 0.31, "learning_rate": 0.0016180182716251034, "loss": 1.666, "step": 3138 }, { "epoch": 0.31, "learning_rate": 0.0016177672180487073, "loss": 1.7305, "step": 3139 }, { "epoch": 0.31, "learning_rate": 0.0016175161014894076, "loss": 1.7275, "step": 3140 }, { "epoch": 0.31, "learning_rate": 0.0016172649219728062, "loss": 1.8066, "step": 3141 }, { "epoch": 0.31, "learning_rate": 0.001617013679524511, "loss": 1.7383, "step": 3142 }, { "epoch": 0.31, "learning_rate": 0.0016167623741701377, "loss": 1.6104, "step": 3143 }, { "epoch": 0.31, "learning_rate": 0.0016165110059353064, "loss": 1.6582, "step": 3144 }, { "epoch": 0.31, "learning_rate": 0.0016162595748456458, "loss": 1.751, "step": 3145 }, { "epoch": 0.31, "learning_rate": 0.0016160080809267895, "loss": 1.668, "step": 3146 }, { "epoch": 0.31, "learning_rate": 0.0016157565242043775, "loss": 1.6914, "step": 3147 }, { "epoch": 0.31, "learning_rate": 0.0016155049047040574, "loss": 1.6318, "step": 3148 }, { "epoch": 0.31, "learning_rate": 0.0016152532224514818, "loss": 1.7559, "step": 3149 }, { "epoch": 0.31, "learning_rate": 0.0016150014774723106, "loss": 1.7441, "step": 3150 }, { "epoch": 0.31, "learning_rate": 0.00161474966979221, "loss": 1.7783, "step": 3151 }, { "epoch": 0.31, "learning_rate": 0.001614497799436852, "loss": 1.6445, "step": 3152 }, { "epoch": 0.31, "learning_rate": 0.0016142458664319154, "loss": 1.6279, "step": 3153 }, { "epoch": 0.31, "learning_rate": 0.001613993870803086, "loss": 1.71, "step": 3154 }, { "epoch": 0.31, "learning_rate": 0.001613741812576055, "loss": 1.6748, "step": 3155 }, { "epoch": 0.31, "learning_rate": 0.0016134896917765205, "loss": 1.6631, "step": 3156 }, { "epoch": 0.31, "learning_rate": 0.0016132375084301862, "loss": 1.584, "step": 3157 }, { "epoch": 0.31, "learning_rate": 0.001612985262562764, "loss": 1.5205, "step": 3158 }, { "epoch": 0.31, "learning_rate": 0.00161273295419997, "loss": 1.7617, "step": 3159 }, { "epoch": 0.31, "learning_rate": 0.0016124805833675278, "loss": 1.7422, "step": 3160 }, { "epoch": 0.31, "learning_rate": 0.0016122281500911678, "loss": 1.7266, "step": 3161 }, { "epoch": 0.31, "learning_rate": 0.0016119756543966255, "loss": 1.6924, "step": 3162 }, { "epoch": 0.31, "learning_rate": 0.0016117230963096439, "loss": 1.6494, "step": 3163 }, { "epoch": 0.31, "learning_rate": 0.0016114704758559716, "loss": 1.7031, "step": 3164 }, { "epoch": 0.31, "learning_rate": 0.0016112177930613645, "loss": 1.6572, "step": 3165 }, { "epoch": 0.31, "learning_rate": 0.001610965047951584, "loss": 1.7256, "step": 3166 }, { "epoch": 0.31, "learning_rate": 0.0016107122405523973, "loss": 1.7666, "step": 3167 }, { "epoch": 0.31, "learning_rate": 0.0016104593708895795, "loss": 1.6592, "step": 3168 }, { "epoch": 0.31, "learning_rate": 0.0016102064389889115, "loss": 1.7275, "step": 3169 }, { "epoch": 0.31, "learning_rate": 0.0016099534448761799, "loss": 1.6572, "step": 3170 }, { "epoch": 0.31, "learning_rate": 0.001609700388577178, "loss": 1.7061, "step": 3171 }, { "epoch": 0.31, "learning_rate": 0.001609447270117706, "loss": 1.6768, "step": 3172 }, { "epoch": 0.31, "learning_rate": 0.001609194089523569, "loss": 1.6904, "step": 3173 }, { "epoch": 0.31, "learning_rate": 0.0016089408468205806, "loss": 1.7148, "step": 3174 }, { "epoch": 0.31, "learning_rate": 0.0016086875420345589, "loss": 1.6885, "step": 3175 }, { "epoch": 0.31, "learning_rate": 0.001608434175191329, "loss": 1.6699, "step": 3176 }, { "epoch": 0.31, "learning_rate": 0.0016081807463167223, "loss": 1.6719, "step": 3177 }, { "epoch": 0.31, "learning_rate": 0.0016079272554365765, "loss": 1.7051, "step": 3178 }, { "epoch": 0.31, "learning_rate": 0.0016076737025767353, "loss": 1.7256, "step": 3179 }, { "epoch": 0.31, "learning_rate": 0.00160742008776305, "loss": 1.7295, "step": 3180 }, { "epoch": 0.31, "learning_rate": 0.001607166411021376, "loss": 1.6592, "step": 3181 }, { "epoch": 0.31, "learning_rate": 0.001606912672377577, "loss": 1.6904, "step": 3182 }, { "epoch": 0.31, "learning_rate": 0.0016066588718575226, "loss": 1.6006, "step": 3183 }, { "epoch": 0.31, "learning_rate": 0.001606405009487088, "loss": 1.6035, "step": 3184 }, { "epoch": 0.31, "learning_rate": 0.0016061510852921547, "loss": 1.6621, "step": 3185 }, { "epoch": 0.31, "learning_rate": 0.0016058970992986112, "loss": 1.7285, "step": 3186 }, { "epoch": 0.31, "learning_rate": 0.0016056430515323527, "loss": 1.6221, "step": 3187 }, { "epoch": 0.31, "learning_rate": 0.0016053889420192791, "loss": 1.6172, "step": 3188 }, { "epoch": 0.31, "learning_rate": 0.0016051347707852976, "loss": 1.4941, "step": 3189 }, { "epoch": 0.31, "learning_rate": 0.0016048805378563221, "loss": 1.5889, "step": 3190 }, { "epoch": 0.31, "learning_rate": 0.0016046262432582717, "loss": 1.7275, "step": 3191 }, { "epoch": 0.31, "learning_rate": 0.0016043718870170727, "loss": 1.6729, "step": 3192 }, { "epoch": 0.31, "learning_rate": 0.0016041174691586574, "loss": 1.7461, "step": 3193 }, { "epoch": 0.31, "learning_rate": 0.001603862989708964, "loss": 1.6826, "step": 3194 }, { "epoch": 0.31, "learning_rate": 0.0016036084486939375, "loss": 1.6836, "step": 3195 }, { "epoch": 0.32, "learning_rate": 0.0016033538461395292, "loss": 1.6855, "step": 3196 }, { "epoch": 0.32, "learning_rate": 0.001603099182071696, "loss": 1.7041, "step": 3197 }, { "epoch": 0.32, "learning_rate": 0.0016028444565164021, "loss": 1.749, "step": 3198 }, { "epoch": 0.32, "learning_rate": 0.0016025896694996172, "loss": 1.6689, "step": 3199 }, { "epoch": 0.32, "learning_rate": 0.001602334821047317, "loss": 1.6445, "step": 3200 }, { "epoch": 0.32, "learning_rate": 0.0016020799111854847, "loss": 1.7197, "step": 3201 }, { "epoch": 0.32, "learning_rate": 0.0016018249399401081, "loss": 1.6377, "step": 3202 }, { "epoch": 0.32, "learning_rate": 0.0016015699073371832, "loss": 1.748, "step": 3203 }, { "epoch": 0.32, "learning_rate": 0.0016013148134027104, "loss": 1.7627, "step": 3204 }, { "epoch": 0.32, "learning_rate": 0.0016010596581626975, "loss": 1.6875, "step": 3205 }, { "epoch": 0.32, "learning_rate": 0.0016008044416431582, "loss": 1.7861, "step": 3206 }, { "epoch": 0.32, "learning_rate": 0.001600549163870112, "loss": 1.6064, "step": 3207 }, { "epoch": 0.32, "learning_rate": 0.0016002938248695863, "loss": 1.6826, "step": 3208 }, { "epoch": 0.32, "learning_rate": 0.001600038424667612, "loss": 1.6689, "step": 3209 }, { "epoch": 0.32, "learning_rate": 0.0015997829632902287, "loss": 1.6973, "step": 3210 }, { "epoch": 0.32, "learning_rate": 0.0015995274407634814, "loss": 1.7197, "step": 3211 }, { "epoch": 0.32, "learning_rate": 0.001599271857113421, "loss": 1.709, "step": 3212 }, { "epoch": 0.32, "learning_rate": 0.0015990162123661047, "loss": 1.873, "step": 3213 }, { "epoch": 0.32, "learning_rate": 0.0015987605065475965, "loss": 1.7246, "step": 3214 }, { "epoch": 0.32, "learning_rate": 0.0015985047396839664, "loss": 1.668, "step": 3215 }, { "epoch": 0.32, "learning_rate": 0.0015982489118012894, "loss": 1.7881, "step": 3216 }, { "epoch": 0.32, "learning_rate": 0.001597993022925649, "loss": 1.7295, "step": 3217 }, { "epoch": 0.32, "learning_rate": 0.001597737073083133, "loss": 1.6572, "step": 3218 }, { "epoch": 0.32, "learning_rate": 0.0015974810622998363, "loss": 1.8037, "step": 3219 }, { "epoch": 0.32, "learning_rate": 0.00159722499060186, "loss": 1.7305, "step": 3220 }, { "epoch": 0.32, "learning_rate": 0.0015969688580153112, "loss": 1.6504, "step": 3221 }, { "epoch": 0.32, "learning_rate": 0.0015967126645663032, "loss": 1.6973, "step": 3222 }, { "epoch": 0.32, "learning_rate": 0.0015964564102809553, "loss": 1.7793, "step": 3223 }, { "epoch": 0.32, "learning_rate": 0.001596200095185394, "loss": 1.6191, "step": 3224 }, { "epoch": 0.32, "learning_rate": 0.0015959437193057504, "loss": 1.667, "step": 3225 }, { "epoch": 0.32, "learning_rate": 0.001595687282668163, "loss": 1.7256, "step": 3226 }, { "epoch": 0.32, "learning_rate": 0.0015954307852987762, "loss": 1.5889, "step": 3227 }, { "epoch": 0.32, "learning_rate": 0.001595174227223741, "loss": 1.6875, "step": 3228 }, { "epoch": 0.32, "learning_rate": 0.001594917608469213, "loss": 1.665, "step": 3229 }, { "epoch": 0.32, "learning_rate": 0.001594660929061356, "loss": 1.7266, "step": 3230 }, { "epoch": 0.32, "learning_rate": 0.001594404189026339, "loss": 1.5107, "step": 3231 }, { "epoch": 0.32, "learning_rate": 0.001594147388390337, "loss": 1.752, "step": 3232 }, { "epoch": 0.32, "learning_rate": 0.0015938905271795317, "loss": 1.624, "step": 3233 }, { "epoch": 0.32, "learning_rate": 0.0015936336054201109, "loss": 1.7246, "step": 3234 }, { "epoch": 0.32, "learning_rate": 0.0015933766231382678, "loss": 1.7666, "step": 3235 }, { "epoch": 0.32, "learning_rate": 0.001593119580360203, "loss": 1.7188, "step": 3236 }, { "epoch": 0.32, "learning_rate": 0.0015928624771121224, "loss": 1.5684, "step": 3237 }, { "epoch": 0.32, "learning_rate": 0.0015926053134202383, "loss": 1.6357, "step": 3238 }, { "epoch": 0.32, "learning_rate": 0.0015923480893107692, "loss": 1.5146, "step": 3239 }, { "epoch": 0.32, "learning_rate": 0.0015920908048099398, "loss": 1.6064, "step": 3240 }, { "epoch": 0.32, "learning_rate": 0.0015918334599439807, "loss": 1.5928, "step": 3241 }, { "epoch": 0.32, "learning_rate": 0.0015915760547391292, "loss": 1.7979, "step": 3242 }, { "epoch": 0.32, "learning_rate": 0.001591318589221628, "loss": 1.6602, "step": 3243 }, { "epoch": 0.32, "learning_rate": 0.0015910610634177265, "loss": 1.7012, "step": 3244 }, { "epoch": 0.32, "learning_rate": 0.0015908034773536805, "loss": 1.6494, "step": 3245 }, { "epoch": 0.32, "learning_rate": 0.0015905458310557508, "loss": 1.7793, "step": 3246 }, { "epoch": 0.32, "learning_rate": 0.0015902881245502055, "loss": 1.8223, "step": 3247 }, { "epoch": 0.32, "learning_rate": 0.0015900303578633185, "loss": 1.6973, "step": 3248 }, { "epoch": 0.32, "learning_rate": 0.0015897725310213696, "loss": 1.665, "step": 3249 }, { "epoch": 0.32, "learning_rate": 0.0015895146440506446, "loss": 1.6982, "step": 3250 }, { "epoch": 0.32, "learning_rate": 0.0015892566969774365, "loss": 1.7285, "step": 3251 }, { "epoch": 0.32, "learning_rate": 0.0015889986898280429, "loss": 1.627, "step": 3252 }, { "epoch": 0.32, "learning_rate": 0.0015887406226287686, "loss": 1.6523, "step": 3253 }, { "epoch": 0.32, "learning_rate": 0.0015884824954059242, "loss": 1.5459, "step": 3254 }, { "epoch": 0.32, "learning_rate": 0.0015882243081858265, "loss": 1.7012, "step": 3255 }, { "epoch": 0.32, "learning_rate": 0.0015879660609947979, "loss": 1.6973, "step": 3256 }, { "epoch": 0.32, "learning_rate": 0.0015877077538591678, "loss": 1.7295, "step": 3257 }, { "epoch": 0.32, "learning_rate": 0.0015874493868052709, "loss": 1.6191, "step": 3258 }, { "epoch": 0.32, "learning_rate": 0.001587190959859449, "loss": 1.6338, "step": 3259 }, { "epoch": 0.32, "learning_rate": 0.0015869324730480483, "loss": 1.7119, "step": 3260 }, { "epoch": 0.32, "learning_rate": 0.001586673926397423, "loss": 1.6719, "step": 3261 }, { "epoch": 0.32, "learning_rate": 0.0015864153199339323, "loss": 1.6709, "step": 3262 }, { "epoch": 0.32, "learning_rate": 0.0015861566536839423, "loss": 1.584, "step": 3263 }, { "epoch": 0.32, "learning_rate": 0.0015858979276738238, "loss": 1.7217, "step": 3264 }, { "epoch": 0.32, "learning_rate": 0.001585639141929955, "loss": 1.7686, "step": 3265 }, { "epoch": 0.32, "learning_rate": 0.0015853802964787196, "loss": 1.6963, "step": 3266 }, { "epoch": 0.32, "learning_rate": 0.001585121391346508, "loss": 1.6748, "step": 3267 }, { "epoch": 0.32, "learning_rate": 0.0015848624265597158, "loss": 1.5791, "step": 3268 }, { "epoch": 0.32, "learning_rate": 0.0015846034021447448, "loss": 1.668, "step": 3269 }, { "epoch": 0.32, "learning_rate": 0.0015843443181280037, "loss": 1.7451, "step": 3270 }, { "epoch": 0.32, "learning_rate": 0.001584085174535907, "loss": 1.6611, "step": 3271 }, { "epoch": 0.32, "learning_rate": 0.0015838259713948743, "loss": 1.7148, "step": 3272 }, { "epoch": 0.32, "learning_rate": 0.0015835667087313323, "loss": 1.623, "step": 3273 }, { "epoch": 0.32, "learning_rate": 0.0015833073865717138, "loss": 1.6504, "step": 3274 }, { "epoch": 0.32, "learning_rate": 0.001583048004942457, "loss": 1.7021, "step": 3275 }, { "epoch": 0.32, "learning_rate": 0.0015827885638700064, "loss": 1.6113, "step": 3276 }, { "epoch": 0.32, "learning_rate": 0.0015825290633808128, "loss": 1.6621, "step": 3277 }, { "epoch": 0.32, "learning_rate": 0.001582269503501333, "loss": 1.5771, "step": 3278 }, { "epoch": 0.32, "learning_rate": 0.0015820098842580298, "loss": 1.6553, "step": 3279 }, { "epoch": 0.32, "learning_rate": 0.001581750205677372, "loss": 1.666, "step": 3280 }, { "epoch": 0.32, "learning_rate": 0.0015814904677858344, "loss": 1.7432, "step": 3281 }, { "epoch": 0.32, "learning_rate": 0.0015812306706098978, "loss": 1.8311, "step": 3282 }, { "epoch": 0.32, "learning_rate": 0.0015809708141760494, "loss": 1.8008, "step": 3283 }, { "epoch": 0.32, "learning_rate": 0.0015807108985107822, "loss": 1.6553, "step": 3284 }, { "epoch": 0.32, "learning_rate": 0.0015804509236405952, "loss": 1.6777, "step": 3285 }, { "epoch": 0.32, "learning_rate": 0.0015801908895919934, "loss": 1.6846, "step": 3286 }, { "epoch": 0.32, "learning_rate": 0.001579930796391488, "loss": 1.5596, "step": 3287 }, { "epoch": 0.32, "learning_rate": 0.001579670644065596, "loss": 1.6309, "step": 3288 }, { "epoch": 0.32, "learning_rate": 0.0015794104326408407, "loss": 1.6807, "step": 3289 }, { "epoch": 0.32, "learning_rate": 0.0015791501621437516, "loss": 1.6865, "step": 3290 }, { "epoch": 0.32, "learning_rate": 0.0015788898326008635, "loss": 1.665, "step": 3291 }, { "epoch": 0.32, "learning_rate": 0.0015786294440387177, "loss": 1.6621, "step": 3292 }, { "epoch": 0.32, "learning_rate": 0.0015783689964838615, "loss": 1.8027, "step": 3293 }, { "epoch": 0.32, "learning_rate": 0.0015781084899628484, "loss": 1.7266, "step": 3294 }, { "epoch": 0.32, "learning_rate": 0.0015778479245022372, "loss": 1.6475, "step": 3295 }, { "epoch": 0.32, "learning_rate": 0.001577587300128594, "loss": 1.6816, "step": 3296 }, { "epoch": 0.33, "learning_rate": 0.001577326616868489, "loss": 1.6982, "step": 3297 }, { "epoch": 0.33, "learning_rate": 0.0015770658747485005, "loss": 1.6357, "step": 3298 }, { "epoch": 0.33, "learning_rate": 0.0015768050737952114, "loss": 1.7188, "step": 3299 }, { "epoch": 0.33, "learning_rate": 0.0015765442140352108, "loss": 1.7959, "step": 3300 }, { "epoch": 0.33, "learning_rate": 0.0015762832954950943, "loss": 1.6592, "step": 3301 }, { "epoch": 0.33, "learning_rate": 0.0015760223182014633, "loss": 1.6504, "step": 3302 }, { "epoch": 0.33, "learning_rate": 0.0015757612821809249, "loss": 1.667, "step": 3303 }, { "epoch": 0.33, "learning_rate": 0.001575500187460092, "loss": 1.751, "step": 3304 }, { "epoch": 0.33, "learning_rate": 0.0015752390340655845, "loss": 1.7695, "step": 3305 }, { "epoch": 0.33, "learning_rate": 0.0015749778220240272, "loss": 1.667, "step": 3306 }, { "epoch": 0.33, "learning_rate": 0.0015747165513620516, "loss": 1.5586, "step": 3307 }, { "epoch": 0.33, "learning_rate": 0.0015744552221062947, "loss": 1.5889, "step": 3308 }, { "epoch": 0.33, "learning_rate": 0.0015741938342834, "loss": 1.7061, "step": 3309 }, { "epoch": 0.33, "learning_rate": 0.0015739323879200163, "loss": 1.625, "step": 3310 }, { "epoch": 0.33, "learning_rate": 0.0015736708830427983, "loss": 1.6572, "step": 3311 }, { "epoch": 0.33, "learning_rate": 0.001573409319678408, "loss": 1.7207, "step": 3312 }, { "epoch": 0.33, "learning_rate": 0.001573147697853512, "loss": 1.6738, "step": 3313 }, { "epoch": 0.33, "learning_rate": 0.0015728860175947832, "loss": 1.6602, "step": 3314 }, { "epoch": 0.33, "learning_rate": 0.0015726242789289007, "loss": 1.6748, "step": 3315 }, { "epoch": 0.33, "learning_rate": 0.0015723624818825492, "loss": 1.6533, "step": 3316 }, { "epoch": 0.33, "learning_rate": 0.0015721006264824199, "loss": 1.7266, "step": 3317 }, { "epoch": 0.33, "learning_rate": 0.0015718387127552091, "loss": 1.7012, "step": 3318 }, { "epoch": 0.33, "learning_rate": 0.0015715767407276204, "loss": 1.6641, "step": 3319 }, { "epoch": 0.33, "learning_rate": 0.0015713147104263617, "loss": 1.6152, "step": 3320 }, { "epoch": 0.33, "learning_rate": 0.0015710526218781476, "loss": 1.7031, "step": 3321 }, { "epoch": 0.33, "learning_rate": 0.0015707904751096993, "loss": 1.5654, "step": 3322 }, { "epoch": 0.33, "learning_rate": 0.0015705282701477432, "loss": 1.7656, "step": 3323 }, { "epoch": 0.33, "learning_rate": 0.0015702660070190113, "loss": 1.6484, "step": 3324 }, { "epoch": 0.33, "learning_rate": 0.0015700036857502421, "loss": 1.6914, "step": 3325 }, { "epoch": 0.33, "learning_rate": 0.0015697413063681803, "loss": 1.6768, "step": 3326 }, { "epoch": 0.33, "learning_rate": 0.001569478868899576, "loss": 1.6719, "step": 3327 }, { "epoch": 0.33, "learning_rate": 0.0015692163733711851, "loss": 1.6807, "step": 3328 }, { "epoch": 0.33, "learning_rate": 0.0015689538198097694, "loss": 1.6035, "step": 3329 }, { "epoch": 0.33, "learning_rate": 0.0015686912082420977, "loss": 1.6729, "step": 3330 }, { "epoch": 0.33, "learning_rate": 0.0015684285386949434, "loss": 1.7305, "step": 3331 }, { "epoch": 0.33, "learning_rate": 0.0015681658111950862, "loss": 1.8242, "step": 3332 }, { "epoch": 0.33, "learning_rate": 0.0015679030257693123, "loss": 1.6162, "step": 3333 }, { "epoch": 0.33, "learning_rate": 0.001567640182444413, "loss": 1.7344, "step": 3334 }, { "epoch": 0.33, "learning_rate": 0.0015673772812471861, "loss": 1.6143, "step": 3335 }, { "epoch": 0.33, "learning_rate": 0.0015671143222044345, "loss": 1.7285, "step": 3336 }, { "epoch": 0.33, "learning_rate": 0.0015668513053429677, "loss": 1.5156, "step": 3337 }, { "epoch": 0.33, "learning_rate": 0.0015665882306896012, "loss": 1.6855, "step": 3338 }, { "epoch": 0.33, "learning_rate": 0.001566325098271156, "loss": 1.6309, "step": 3339 }, { "epoch": 0.33, "learning_rate": 0.0015660619081144592, "loss": 1.6924, "step": 3340 }, { "epoch": 0.33, "learning_rate": 0.0015657986602463432, "loss": 1.5947, "step": 3341 }, { "epoch": 0.33, "learning_rate": 0.0015655353546936476, "loss": 1.624, "step": 3342 }, { "epoch": 0.33, "learning_rate": 0.001565271991483216, "loss": 1.7393, "step": 3343 }, { "epoch": 0.33, "learning_rate": 0.0015650085706419002, "loss": 1.6836, "step": 3344 }, { "epoch": 0.33, "learning_rate": 0.0015647450921965555, "loss": 1.6826, "step": 3345 }, { "epoch": 0.33, "learning_rate": 0.001564481556174045, "loss": 1.7344, "step": 3346 }, { "epoch": 0.33, "learning_rate": 0.001564217962601236, "loss": 1.5762, "step": 3347 }, { "epoch": 0.33, "learning_rate": 0.0015639543115050034, "loss": 1.6992, "step": 3348 }, { "epoch": 0.33, "learning_rate": 0.0015636906029122266, "loss": 1.6289, "step": 3349 }, { "epoch": 0.33, "learning_rate": 0.001563426836849791, "loss": 1.7539, "step": 3350 }, { "epoch": 0.33, "learning_rate": 0.0015631630133445893, "loss": 1.7178, "step": 3351 }, { "epoch": 0.33, "learning_rate": 0.001562899132423518, "loss": 1.7139, "step": 3352 }, { "epoch": 0.33, "learning_rate": 0.0015626351941134807, "loss": 1.7744, "step": 3353 }, { "epoch": 0.33, "learning_rate": 0.0015623711984413867, "loss": 1.6514, "step": 3354 }, { "epoch": 0.33, "learning_rate": 0.0015621071454341507, "loss": 1.6934, "step": 3355 }, { "epoch": 0.33, "learning_rate": 0.0015618430351186938, "loss": 1.5977, "step": 3356 }, { "epoch": 0.33, "learning_rate": 0.0015615788675219426, "loss": 1.7949, "step": 3357 }, { "epoch": 0.33, "learning_rate": 0.00156131464267083, "loss": 1.7305, "step": 3358 }, { "epoch": 0.33, "learning_rate": 0.0015610503605922941, "loss": 1.6836, "step": 3359 }, { "epoch": 0.33, "learning_rate": 0.0015607860213132788, "loss": 1.6299, "step": 3360 }, { "epoch": 0.33, "learning_rate": 0.0015605216248607346, "loss": 1.6768, "step": 3361 }, { "epoch": 0.33, "learning_rate": 0.0015602571712616169, "loss": 1.7002, "step": 3362 }, { "epoch": 0.33, "learning_rate": 0.0015599926605428882, "loss": 1.7178, "step": 3363 }, { "epoch": 0.33, "learning_rate": 0.0015597280927315152, "loss": 1.6201, "step": 3364 }, { "epoch": 0.33, "learning_rate": 0.0015594634678544716, "loss": 1.5713, "step": 3365 }, { "epoch": 0.33, "learning_rate": 0.0015591987859387368, "loss": 1.6221, "step": 3366 }, { "epoch": 0.33, "learning_rate": 0.0015589340470112956, "loss": 1.6113, "step": 3367 }, { "epoch": 0.33, "learning_rate": 0.0015586692510991384, "loss": 1.7119, "step": 3368 }, { "epoch": 0.33, "learning_rate": 0.0015584043982292625, "loss": 1.6309, "step": 3369 }, { "epoch": 0.33, "learning_rate": 0.0015581394884286697, "loss": 1.7119, "step": 3370 }, { "epoch": 0.33, "learning_rate": 0.0015578745217243682, "loss": 1.6367, "step": 3371 }, { "epoch": 0.33, "learning_rate": 0.0015576094981433726, "loss": 1.6377, "step": 3372 }, { "epoch": 0.33, "learning_rate": 0.0015573444177127023, "loss": 1.6328, "step": 3373 }, { "epoch": 0.33, "learning_rate": 0.0015570792804593832, "loss": 1.6426, "step": 3374 }, { "epoch": 0.33, "learning_rate": 0.0015568140864104462, "loss": 1.666, "step": 3375 }, { "epoch": 0.33, "learning_rate": 0.0015565488355929289, "loss": 1.7744, "step": 3376 }, { "epoch": 0.33, "learning_rate": 0.001556283528033874, "loss": 1.5781, "step": 3377 }, { "epoch": 0.33, "learning_rate": 0.0015560181637603305, "loss": 1.5889, "step": 3378 }, { "epoch": 0.33, "learning_rate": 0.0015557527427993531, "loss": 1.5869, "step": 3379 }, { "epoch": 0.33, "learning_rate": 0.0015554872651780017, "loss": 1.6025, "step": 3380 }, { "epoch": 0.33, "learning_rate": 0.0015552217309233428, "loss": 1.7246, "step": 3381 }, { "epoch": 0.33, "learning_rate": 0.0015549561400624477, "loss": 1.6289, "step": 3382 }, { "epoch": 0.33, "learning_rate": 0.0015546904926223948, "loss": 1.5645, "step": 3383 }, { "epoch": 0.33, "learning_rate": 0.0015544247886302672, "loss": 1.7451, "step": 3384 }, { "epoch": 0.33, "learning_rate": 0.001554159028113154, "loss": 1.6982, "step": 3385 }, { "epoch": 0.33, "learning_rate": 0.0015538932110981498, "loss": 1.7402, "step": 3386 }, { "epoch": 0.33, "learning_rate": 0.001553627337612356, "loss": 1.6465, "step": 3387 }, { "epoch": 0.33, "learning_rate": 0.0015533614076828786, "loss": 1.6611, "step": 3388 }, { "epoch": 0.33, "learning_rate": 0.00155309542133683, "loss": 1.6973, "step": 3389 }, { "epoch": 0.33, "learning_rate": 0.0015528293786013282, "loss": 1.6992, "step": 3390 }, { "epoch": 0.33, "learning_rate": 0.0015525632795034966, "loss": 1.7461, "step": 3391 }, { "epoch": 0.33, "learning_rate": 0.0015522971240704653, "loss": 1.6865, "step": 3392 }, { "epoch": 0.33, "learning_rate": 0.0015520309123293688, "loss": 1.665, "step": 3393 }, { "epoch": 0.33, "learning_rate": 0.0015517646443073488, "loss": 1.627, "step": 3394 }, { "epoch": 0.33, "learning_rate": 0.001551498320031551, "loss": 1.8984, "step": 3395 }, { "epoch": 0.33, "learning_rate": 0.0015512319395291287, "loss": 1.7227, "step": 3396 }, { "epoch": 0.33, "learning_rate": 0.0015509655028272396, "loss": 1.5684, "step": 3397 }, { "epoch": 0.33, "learning_rate": 0.001550699009953048, "loss": 1.6924, "step": 3398 }, { "epoch": 0.34, "learning_rate": 0.001550432460933723, "loss": 1.6699, "step": 3399 }, { "epoch": 0.34, "learning_rate": 0.0015501658557964403, "loss": 1.7861, "step": 3400 }, { "epoch": 0.34, "learning_rate": 0.0015498991945683805, "loss": 1.6562, "step": 3401 }, { "epoch": 0.34, "learning_rate": 0.001549632477276731, "loss": 1.7344, "step": 3402 }, { "epoch": 0.34, "learning_rate": 0.0015493657039486839, "loss": 1.6543, "step": 3403 }, { "epoch": 0.34, "learning_rate": 0.0015490988746114374, "loss": 1.624, "step": 3404 }, { "epoch": 0.34, "learning_rate": 0.0015488319892921956, "loss": 1.7861, "step": 3405 }, { "epoch": 0.34, "learning_rate": 0.001548565048018168, "loss": 1.6133, "step": 3406 }, { "epoch": 0.34, "learning_rate": 0.00154829805081657, "loss": 1.6934, "step": 3407 }, { "epoch": 0.34, "learning_rate": 0.0015480309977146227, "loss": 1.6641, "step": 3408 }, { "epoch": 0.34, "learning_rate": 0.0015477638887395524, "loss": 1.4609, "step": 3409 }, { "epoch": 0.34, "learning_rate": 0.0015474967239185922, "loss": 1.6621, "step": 3410 }, { "epoch": 0.34, "learning_rate": 0.0015472295032789797, "loss": 1.7061, "step": 3411 }, { "epoch": 0.34, "learning_rate": 0.0015469622268479589, "loss": 1.709, "step": 3412 }, { "epoch": 0.34, "learning_rate": 0.0015466948946527793, "loss": 1.7588, "step": 3413 }, { "epoch": 0.34, "learning_rate": 0.001546427506720696, "loss": 1.666, "step": 3414 }, { "epoch": 0.34, "learning_rate": 0.0015461600630789698, "loss": 1.7568, "step": 3415 }, { "epoch": 0.34, "learning_rate": 0.0015458925637548675, "loss": 1.6436, "step": 3416 }, { "epoch": 0.34, "learning_rate": 0.0015456250087756609, "loss": 1.543, "step": 3417 }, { "epoch": 0.34, "learning_rate": 0.0015453573981686285, "loss": 1.6182, "step": 3418 }, { "epoch": 0.34, "learning_rate": 0.001545089731961053, "loss": 1.6309, "step": 3419 }, { "epoch": 0.34, "learning_rate": 0.0015448220101802247, "loss": 1.7715, "step": 3420 }, { "epoch": 0.34, "learning_rate": 0.0015445542328534376, "loss": 1.6406, "step": 3421 }, { "epoch": 0.34, "learning_rate": 0.001544286400007993, "loss": 1.6973, "step": 3422 }, { "epoch": 0.34, "learning_rate": 0.0015440185116711966, "loss": 1.6455, "step": 3423 }, { "epoch": 0.34, "learning_rate": 0.0015437505678703605, "loss": 1.6143, "step": 3424 }, { "epoch": 0.34, "learning_rate": 0.0015434825686328022, "loss": 1.668, "step": 3425 }, { "epoch": 0.34, "learning_rate": 0.0015432145139858446, "loss": 1.5586, "step": 3426 }, { "epoch": 0.34, "learning_rate": 0.0015429464039568171, "loss": 1.6074, "step": 3427 }, { "epoch": 0.34, "learning_rate": 0.0015426782385730537, "loss": 1.6807, "step": 3428 }, { "epoch": 0.34, "learning_rate": 0.0015424100178618948, "loss": 1.6758, "step": 3429 }, { "epoch": 0.34, "learning_rate": 0.0015421417418506858, "loss": 1.667, "step": 3430 }, { "epoch": 0.34, "learning_rate": 0.0015418734105667787, "loss": 1.6562, "step": 3431 }, { "epoch": 0.34, "learning_rate": 0.0015416050240375304, "loss": 1.7686, "step": 3432 }, { "epoch": 0.34, "learning_rate": 0.001541336582290303, "loss": 1.6221, "step": 3433 }, { "epoch": 0.34, "learning_rate": 0.0015410680853524654, "loss": 1.7051, "step": 3434 }, { "epoch": 0.34, "learning_rate": 0.0015407995332513914, "loss": 1.584, "step": 3435 }, { "epoch": 0.34, "learning_rate": 0.0015405309260144602, "loss": 1.6221, "step": 3436 }, { "epoch": 0.34, "learning_rate": 0.0015402622636690574, "loss": 1.7432, "step": 3437 }, { "epoch": 0.34, "learning_rate": 0.0015399935462425739, "loss": 1.6895, "step": 3438 }, { "epoch": 0.34, "learning_rate": 0.0015397247737624058, "loss": 1.7529, "step": 3439 }, { "epoch": 0.34, "learning_rate": 0.0015394559462559552, "loss": 1.6631, "step": 3440 }, { "epoch": 0.34, "learning_rate": 0.0015391870637506294, "loss": 1.7129, "step": 3441 }, { "epoch": 0.34, "learning_rate": 0.0015389181262738423, "loss": 1.6299, "step": 3442 }, { "epoch": 0.34, "learning_rate": 0.0015386491338530122, "loss": 1.6426, "step": 3443 }, { "epoch": 0.34, "learning_rate": 0.001538380086515564, "loss": 1.5918, "step": 3444 }, { "epoch": 0.34, "learning_rate": 0.0015381109842889274, "loss": 1.7002, "step": 3445 }, { "epoch": 0.34, "learning_rate": 0.0015378418272005382, "loss": 1.5674, "step": 3446 }, { "epoch": 0.34, "learning_rate": 0.0015375726152778372, "loss": 1.6455, "step": 3447 }, { "epoch": 0.34, "learning_rate": 0.001537303348548272, "loss": 1.5059, "step": 3448 }, { "epoch": 0.34, "learning_rate": 0.0015370340270392945, "loss": 1.6562, "step": 3449 }, { "epoch": 0.34, "learning_rate": 0.0015367646507783625, "loss": 1.7783, "step": 3450 }, { "epoch": 0.34, "learning_rate": 0.0015364952197929403, "loss": 1.5088, "step": 3451 }, { "epoch": 0.34, "learning_rate": 0.0015362257341104968, "loss": 1.79, "step": 3452 }, { "epoch": 0.34, "learning_rate": 0.0015359561937585062, "loss": 1.5537, "step": 3453 }, { "epoch": 0.34, "learning_rate": 0.0015356865987644493, "loss": 1.5967, "step": 3454 }, { "epoch": 0.34, "learning_rate": 0.0015354169491558118, "loss": 1.6445, "step": 3455 }, { "epoch": 0.34, "learning_rate": 0.0015351472449600854, "loss": 1.6074, "step": 3456 }, { "epoch": 0.34, "learning_rate": 0.001534877486204767, "loss": 1.6924, "step": 3457 }, { "epoch": 0.34, "learning_rate": 0.0015346076729173589, "loss": 1.7119, "step": 3458 }, { "epoch": 0.34, "learning_rate": 0.0015343378051253694, "loss": 1.6904, "step": 3459 }, { "epoch": 0.34, "learning_rate": 0.0015340678828563124, "loss": 1.6396, "step": 3460 }, { "epoch": 0.34, "learning_rate": 0.001533797906137707, "loss": 1.7314, "step": 3461 }, { "epoch": 0.34, "learning_rate": 0.0015335278749970779, "loss": 1.6943, "step": 3462 }, { "epoch": 0.34, "learning_rate": 0.0015332577894619558, "loss": 1.459, "step": 3463 }, { "epoch": 0.34, "learning_rate": 0.0015329876495598757, "loss": 1.6699, "step": 3464 }, { "epoch": 0.34, "learning_rate": 0.0015327174553183803, "loss": 1.666, "step": 3465 }, { "epoch": 0.34, "learning_rate": 0.0015324472067650152, "loss": 1.6104, "step": 3466 }, { "epoch": 0.34, "learning_rate": 0.0015321769039273344, "loss": 1.6162, "step": 3467 }, { "epoch": 0.34, "learning_rate": 0.001531906546832895, "loss": 1.5693, "step": 3468 }, { "epoch": 0.34, "learning_rate": 0.0015316361355092606, "loss": 1.6836, "step": 3469 }, { "epoch": 0.34, "learning_rate": 0.0015313656699840006, "loss": 1.6289, "step": 3470 }, { "epoch": 0.34, "learning_rate": 0.0015310951502846893, "loss": 1.6035, "step": 3471 }, { "epoch": 0.34, "learning_rate": 0.0015308245764389073, "loss": 1.7383, "step": 3472 }, { "epoch": 0.34, "learning_rate": 0.00153055394847424, "loss": 1.6348, "step": 3473 }, { "epoch": 0.34, "learning_rate": 0.0015302832664182784, "loss": 1.6885, "step": 3474 }, { "epoch": 0.34, "learning_rate": 0.0015300125302986193, "loss": 1.5605, "step": 3475 }, { "epoch": 0.34, "learning_rate": 0.0015297417401428657, "loss": 1.5664, "step": 3476 }, { "epoch": 0.34, "learning_rate": 0.0015294708959786244, "loss": 1.6328, "step": 3477 }, { "epoch": 0.34, "learning_rate": 0.001529199997833509, "loss": 1.7021, "step": 3478 }, { "epoch": 0.34, "learning_rate": 0.001528929045735138, "loss": 1.7139, "step": 3479 }, { "epoch": 0.34, "learning_rate": 0.0015286580397111359, "loss": 1.6582, "step": 3480 }, { "epoch": 0.34, "learning_rate": 0.001528386979789132, "loss": 1.6025, "step": 3481 }, { "epoch": 0.34, "learning_rate": 0.0015281158659967624, "loss": 1.6904, "step": 3482 }, { "epoch": 0.34, "learning_rate": 0.0015278446983616667, "loss": 1.7119, "step": 3483 }, { "epoch": 0.34, "learning_rate": 0.0015275734769114922, "loss": 1.624, "step": 3484 }, { "epoch": 0.34, "learning_rate": 0.00152730220167389, "loss": 1.7139, "step": 3485 }, { "epoch": 0.34, "learning_rate": 0.0015270308726765174, "loss": 1.6367, "step": 3486 }, { "epoch": 0.34, "learning_rate": 0.001526759489947037, "loss": 1.6309, "step": 3487 }, { "epoch": 0.34, "learning_rate": 0.0015264880535131172, "loss": 1.5859, "step": 3488 }, { "epoch": 0.34, "learning_rate": 0.0015262165634024314, "loss": 1.5908, "step": 3489 }, { "epoch": 0.34, "learning_rate": 0.0015259450196426587, "loss": 1.5889, "step": 3490 }, { "epoch": 0.34, "learning_rate": 0.0015256734222614833, "loss": 1.6006, "step": 3491 }, { "epoch": 0.34, "learning_rate": 0.001525401771286596, "loss": 1.71, "step": 3492 }, { "epoch": 0.34, "learning_rate": 0.001525130066745692, "loss": 1.6387, "step": 3493 }, { "epoch": 0.34, "learning_rate": 0.001524858308666472, "loss": 1.7441, "step": 3494 }, { "epoch": 0.34, "learning_rate": 0.0015245864970766423, "loss": 1.7305, "step": 3495 }, { "epoch": 0.34, "learning_rate": 0.001524314632003916, "loss": 1.6553, "step": 3496 }, { "epoch": 0.34, "learning_rate": 0.0015240427134760085, "loss": 1.6816, "step": 3497 }, { "epoch": 0.34, "learning_rate": 0.001523770741520644, "loss": 1.7061, "step": 3498 }, { "epoch": 0.34, "learning_rate": 0.0015234987161655495, "loss": 1.6201, "step": 3499 }, { "epoch": 0.35, "learning_rate": 0.00152322663743846, "loss": 1.7246, "step": 3500 }, { "epoch": 0.35, "learning_rate": 0.0015229545053671138, "loss": 1.6523, "step": 3501 }, { "epoch": 0.35, "learning_rate": 0.001522682319979256, "loss": 1.5117, "step": 3502 }, { "epoch": 0.35, "learning_rate": 0.001522410081302636, "loss": 1.5518, "step": 3503 }, { "epoch": 0.35, "learning_rate": 0.0015221377893650098, "loss": 1.7148, "step": 3504 }, { "epoch": 0.35, "learning_rate": 0.0015218654441941373, "loss": 1.5146, "step": 3505 }, { "epoch": 0.35, "learning_rate": 0.0015215930458177857, "loss": 1.6504, "step": 3506 }, { "epoch": 0.35, "learning_rate": 0.001521320594263726, "loss": 1.7031, "step": 3507 }, { "epoch": 0.35, "learning_rate": 0.0015210480895597358, "loss": 1.5957, "step": 3508 }, { "epoch": 0.35, "learning_rate": 0.0015207755317335975, "loss": 1.7441, "step": 3509 }, { "epoch": 0.35, "learning_rate": 0.0015205029208130991, "loss": 1.6963, "step": 3510 }, { "epoch": 0.35, "learning_rate": 0.001520230256826034, "loss": 1.6045, "step": 3511 }, { "epoch": 0.35, "learning_rate": 0.0015199575398002008, "loss": 1.5664, "step": 3512 }, { "epoch": 0.35, "learning_rate": 0.0015196847697634035, "loss": 1.5713, "step": 3513 }, { "epoch": 0.35, "learning_rate": 0.001519411946743452, "loss": 1.8135, "step": 3514 }, { "epoch": 0.35, "learning_rate": 0.0015191390707681616, "loss": 1.6729, "step": 3515 }, { "epoch": 0.35, "learning_rate": 0.001518866141865352, "loss": 1.7656, "step": 3516 }, { "epoch": 0.35, "learning_rate": 0.0015185931600628488, "loss": 1.6016, "step": 3517 }, { "epoch": 0.35, "learning_rate": 0.0015183201253884842, "loss": 1.5996, "step": 3518 }, { "epoch": 0.35, "learning_rate": 0.0015180470378700944, "loss": 1.5273, "step": 3519 }, { "epoch": 0.35, "learning_rate": 0.0015177738975355206, "loss": 1.584, "step": 3520 }, { "epoch": 0.35, "learning_rate": 0.0015175007044126111, "loss": 1.8008, "step": 3521 }, { "epoch": 0.35, "learning_rate": 0.001517227458529218, "loss": 1.5879, "step": 3522 }, { "epoch": 0.35, "learning_rate": 0.0015169541599131997, "loss": 1.8096, "step": 3523 }, { "epoch": 0.35, "learning_rate": 0.0015166808085924197, "loss": 1.7021, "step": 3524 }, { "epoch": 0.35, "learning_rate": 0.0015164074045947466, "loss": 1.5625, "step": 3525 }, { "epoch": 0.35, "learning_rate": 0.0015161339479480547, "loss": 1.6748, "step": 3526 }, { "epoch": 0.35, "learning_rate": 0.0015158604386802238, "loss": 1.707, "step": 3527 }, { "epoch": 0.35, "learning_rate": 0.0015155868768191385, "loss": 1.5898, "step": 3528 }, { "epoch": 0.35, "learning_rate": 0.0015153132623926895, "loss": 1.6523, "step": 3529 }, { "epoch": 0.35, "learning_rate": 0.0015150395954287724, "loss": 1.6299, "step": 3530 }, { "epoch": 0.35, "learning_rate": 0.0015147658759552876, "loss": 1.709, "step": 3531 }, { "epoch": 0.35, "learning_rate": 0.0015144921040001424, "loss": 1.5908, "step": 3532 }, { "epoch": 0.35, "learning_rate": 0.0015142182795912476, "loss": 1.6191, "step": 3533 }, { "epoch": 0.35, "learning_rate": 0.0015139444027565212, "loss": 1.7002, "step": 3534 }, { "epoch": 0.35, "learning_rate": 0.0015136704735238852, "loss": 1.7021, "step": 3535 }, { "epoch": 0.35, "learning_rate": 0.0015133964919212674, "loss": 1.6133, "step": 3536 }, { "epoch": 0.35, "learning_rate": 0.0015131224579766006, "loss": 1.7383, "step": 3537 }, { "epoch": 0.35, "learning_rate": 0.0015128483717178236, "loss": 1.6494, "step": 3538 }, { "epoch": 0.35, "learning_rate": 0.0015125742331728803, "loss": 1.6221, "step": 3539 }, { "epoch": 0.35, "learning_rate": 0.0015123000423697196, "loss": 1.7607, "step": 3540 }, { "epoch": 0.35, "learning_rate": 0.001512025799336296, "loss": 1.6846, "step": 3541 }, { "epoch": 0.35, "learning_rate": 0.0015117515041005688, "loss": 1.6523, "step": 3542 }, { "epoch": 0.35, "learning_rate": 0.0015114771566905039, "loss": 1.6143, "step": 3543 }, { "epoch": 0.35, "learning_rate": 0.001511202757134071, "loss": 1.8037, "step": 3544 }, { "epoch": 0.35, "learning_rate": 0.0015109283054592461, "loss": 1.6387, "step": 3545 }, { "epoch": 0.35, "learning_rate": 0.0015106538016940103, "loss": 1.6338, "step": 3546 }, { "epoch": 0.35, "learning_rate": 0.0015103792458663501, "loss": 1.5967, "step": 3547 }, { "epoch": 0.35, "learning_rate": 0.0015101046380042566, "loss": 1.6543, "step": 3548 }, { "epoch": 0.35, "learning_rate": 0.0015098299781357274, "loss": 1.5625, "step": 3549 }, { "epoch": 0.35, "learning_rate": 0.0015095552662887641, "loss": 1.5107, "step": 3550 }, { "epoch": 0.35, "learning_rate": 0.0015092805024913747, "loss": 1.5547, "step": 3551 }, { "epoch": 0.35, "learning_rate": 0.0015090056867715722, "loss": 1.6943, "step": 3552 }, { "epoch": 0.35, "learning_rate": 0.0015087308191573743, "loss": 1.6162, "step": 3553 }, { "epoch": 0.35, "learning_rate": 0.0015084558996768045, "loss": 1.6533, "step": 3554 }, { "epoch": 0.35, "learning_rate": 0.0015081809283578919, "loss": 1.7422, "step": 3555 }, { "epoch": 0.35, "learning_rate": 0.00150790590522867, "loss": 1.6836, "step": 3556 }, { "epoch": 0.35, "learning_rate": 0.0015076308303171783, "loss": 1.6914, "step": 3557 }, { "epoch": 0.35, "learning_rate": 0.0015073557036514613, "loss": 1.6738, "step": 3558 }, { "epoch": 0.35, "learning_rate": 0.0015070805252595694, "loss": 1.7021, "step": 3559 }, { "epoch": 0.35, "learning_rate": 0.001506805295169557, "loss": 1.5957, "step": 3560 }, { "epoch": 0.35, "learning_rate": 0.0015065300134094847, "loss": 1.6758, "step": 3561 }, { "epoch": 0.35, "learning_rate": 0.0015062546800074187, "loss": 1.5801, "step": 3562 }, { "epoch": 0.35, "learning_rate": 0.001505979294991429, "loss": 1.6641, "step": 3563 }, { "epoch": 0.35, "learning_rate": 0.0015057038583895924, "loss": 1.6475, "step": 3564 }, { "epoch": 0.35, "learning_rate": 0.0015054283702299902, "loss": 1.6895, "step": 3565 }, { "epoch": 0.35, "learning_rate": 0.0015051528305407092, "loss": 1.6221, "step": 3566 }, { "epoch": 0.35, "learning_rate": 0.001504877239349841, "loss": 1.5898, "step": 3567 }, { "epoch": 0.35, "learning_rate": 0.0015046015966854833, "loss": 1.6523, "step": 3568 }, { "epoch": 0.35, "learning_rate": 0.0015043259025757383, "loss": 1.7207, "step": 3569 }, { "epoch": 0.35, "learning_rate": 0.0015040501570487135, "loss": 1.6152, "step": 3570 }, { "epoch": 0.35, "learning_rate": 0.0015037743601325222, "loss": 1.6143, "step": 3571 }, { "epoch": 0.35, "learning_rate": 0.0015034985118552826, "loss": 1.583, "step": 3572 }, { "epoch": 0.35, "learning_rate": 0.001503222612245118, "loss": 1.6367, "step": 3573 }, { "epoch": 0.35, "learning_rate": 0.0015029466613301567, "loss": 1.6719, "step": 3574 }, { "epoch": 0.35, "learning_rate": 0.0015026706591385328, "loss": 1.6797, "step": 3575 }, { "epoch": 0.35, "learning_rate": 0.0015023946056983855, "loss": 1.6475, "step": 3576 }, { "epoch": 0.35, "learning_rate": 0.0015021185010378596, "loss": 1.6602, "step": 3577 }, { "epoch": 0.35, "learning_rate": 0.001501842345185104, "loss": 1.7803, "step": 3578 }, { "epoch": 0.35, "learning_rate": 0.0015015661381682734, "loss": 1.5879, "step": 3579 }, { "epoch": 0.35, "learning_rate": 0.0015012898800155286, "loss": 1.7988, "step": 3580 }, { "epoch": 0.35, "learning_rate": 0.0015010135707550338, "loss": 1.582, "step": 3581 }, { "epoch": 0.35, "learning_rate": 0.0015007372104149602, "loss": 1.7041, "step": 3582 }, { "epoch": 0.35, "learning_rate": 0.0015004607990234827, "loss": 1.499, "step": 3583 }, { "epoch": 0.35, "learning_rate": 0.0015001843366087826, "loss": 1.6807, "step": 3584 }, { "epoch": 0.35, "learning_rate": 0.0014999078231990462, "loss": 1.665, "step": 3585 }, { "epoch": 0.35, "learning_rate": 0.001499631258822464, "loss": 1.7285, "step": 3586 }, { "epoch": 0.35, "learning_rate": 0.0014993546435072332, "loss": 1.6445, "step": 3587 }, { "epoch": 0.35, "learning_rate": 0.0014990779772815549, "loss": 1.6826, "step": 3588 }, { "epoch": 0.35, "learning_rate": 0.001498801260173636, "loss": 1.6035, "step": 3589 }, { "epoch": 0.35, "learning_rate": 0.0014985244922116885, "loss": 1.5537, "step": 3590 }, { "epoch": 0.35, "learning_rate": 0.0014982476734239295, "loss": 1.6143, "step": 3591 }, { "epoch": 0.35, "learning_rate": 0.0014979708038385818, "loss": 1.6377, "step": 3592 }, { "epoch": 0.35, "learning_rate": 0.0014976938834838725, "loss": 1.5771, "step": 3593 }, { "epoch": 0.35, "learning_rate": 0.0014974169123880344, "loss": 1.6826, "step": 3594 }, { "epoch": 0.35, "learning_rate": 0.0014971398905793054, "loss": 1.6768, "step": 3595 }, { "epoch": 0.35, "learning_rate": 0.001496862818085929, "loss": 1.7461, "step": 3596 }, { "epoch": 0.35, "learning_rate": 0.0014965856949361526, "loss": 1.7197, "step": 3597 }, { "epoch": 0.35, "learning_rate": 0.00149630852115823, "loss": 1.7773, "step": 3598 }, { "epoch": 0.35, "learning_rate": 0.0014960312967804202, "loss": 1.5723, "step": 3599 }, { "epoch": 0.35, "learning_rate": 0.0014957540218309862, "loss": 1.7217, "step": 3600 }, { "epoch": 0.35, "learning_rate": 0.0014954766963381969, "loss": 1.624, "step": 3601 }, { "epoch": 0.36, "learning_rate": 0.0014951993203303272, "loss": 1.6338, "step": 3602 }, { "epoch": 0.36, "learning_rate": 0.0014949218938356552, "loss": 1.6436, "step": 3603 }, { "epoch": 0.36, "learning_rate": 0.001494644416882466, "loss": 1.7266, "step": 3604 }, { "epoch": 0.36, "learning_rate": 0.0014943668894990484, "loss": 1.5947, "step": 3605 }, { "epoch": 0.36, "learning_rate": 0.0014940893117136977, "loss": 1.625, "step": 3606 }, { "epoch": 0.36, "learning_rate": 0.0014938116835547133, "loss": 1.6758, "step": 3607 }, { "epoch": 0.36, "learning_rate": 0.0014935340050504001, "loss": 1.5723, "step": 3608 }, { "epoch": 0.36, "learning_rate": 0.0014932562762290677, "loss": 1.6035, "step": 3609 }, { "epoch": 0.36, "learning_rate": 0.0014929784971190322, "loss": 1.7734, "step": 3610 }, { "epoch": 0.36, "learning_rate": 0.001492700667748613, "loss": 1.7354, "step": 3611 }, { "epoch": 0.36, "learning_rate": 0.0014924227881461358, "loss": 1.7568, "step": 3612 }, { "epoch": 0.36, "learning_rate": 0.0014921448583399312, "loss": 1.7939, "step": 3613 }, { "epoch": 0.36, "learning_rate": 0.0014918668783583347, "loss": 1.6885, "step": 3614 }, { "epoch": 0.36, "learning_rate": 0.001491588848229687, "loss": 1.5732, "step": 3615 }, { "epoch": 0.36, "learning_rate": 0.0014913107679823344, "loss": 1.6455, "step": 3616 }, { "epoch": 0.36, "learning_rate": 0.001491032637644627, "loss": 1.5977, "step": 3617 }, { "epoch": 0.36, "learning_rate": 0.0014907544572449218, "loss": 1.585, "step": 3618 }, { "epoch": 0.36, "learning_rate": 0.0014904762268115796, "loss": 1.6211, "step": 3619 }, { "epoch": 0.36, "learning_rate": 0.001490197946372967, "loss": 1.624, "step": 3620 }, { "epoch": 0.36, "learning_rate": 0.0014899196159574547, "loss": 1.5117, "step": 3621 }, { "epoch": 0.36, "learning_rate": 0.0014896412355934194, "loss": 1.4971, "step": 3622 }, { "epoch": 0.36, "learning_rate": 0.0014893628053092433, "loss": 1.5576, "step": 3623 }, { "epoch": 0.36, "learning_rate": 0.001489084325133312, "loss": 1.5898, "step": 3624 }, { "epoch": 0.36, "learning_rate": 0.0014888057950940183, "loss": 1.6953, "step": 3625 }, { "epoch": 0.36, "learning_rate": 0.0014885272152197582, "loss": 1.6064, "step": 3626 }, { "epoch": 0.36, "learning_rate": 0.0014882485855389344, "loss": 1.6475, "step": 3627 }, { "epoch": 0.36, "learning_rate": 0.001487969906079953, "loss": 1.6025, "step": 3628 }, { "epoch": 0.36, "learning_rate": 0.0014876911768712268, "loss": 1.5742, "step": 3629 }, { "epoch": 0.36, "learning_rate": 0.0014874123979411727, "loss": 1.6934, "step": 3630 }, { "epoch": 0.36, "learning_rate": 0.0014871335693182127, "loss": 1.6729, "step": 3631 }, { "epoch": 0.36, "learning_rate": 0.0014868546910307744, "loss": 1.6758, "step": 3632 }, { "epoch": 0.36, "learning_rate": 0.0014865757631072898, "loss": 1.6475, "step": 3633 }, { "epoch": 0.36, "learning_rate": 0.0014862967855761966, "loss": 1.6758, "step": 3634 }, { "epoch": 0.36, "learning_rate": 0.0014860177584659368, "loss": 1.6338, "step": 3635 }, { "epoch": 0.36, "learning_rate": 0.0014857386818049587, "loss": 1.6162, "step": 3636 }, { "epoch": 0.36, "learning_rate": 0.0014854595556217144, "loss": 1.5615, "step": 3637 }, { "epoch": 0.36, "learning_rate": 0.0014851803799446612, "loss": 1.6582, "step": 3638 }, { "epoch": 0.36, "learning_rate": 0.0014849011548022622, "loss": 1.6611, "step": 3639 }, { "epoch": 0.36, "learning_rate": 0.0014846218802229848, "loss": 1.6113, "step": 3640 }, { "epoch": 0.36, "learning_rate": 0.001484342556235302, "loss": 1.6475, "step": 3641 }, { "epoch": 0.36, "learning_rate": 0.0014840631828676911, "loss": 1.6025, "step": 3642 }, { "epoch": 0.36, "learning_rate": 0.0014837837601486357, "loss": 1.5449, "step": 3643 }, { "epoch": 0.36, "learning_rate": 0.001483504288106623, "loss": 1.6299, "step": 3644 }, { "epoch": 0.36, "learning_rate": 0.001483224766770146, "loss": 1.6436, "step": 3645 }, { "epoch": 0.36, "learning_rate": 0.001482945196167703, "loss": 1.6631, "step": 3646 }, { "epoch": 0.36, "learning_rate": 0.0014826655763277964, "loss": 1.6045, "step": 3647 }, { "epoch": 0.36, "learning_rate": 0.0014823859072789346, "loss": 1.5713, "step": 3648 }, { "epoch": 0.36, "learning_rate": 0.0014821061890496299, "loss": 1.5947, "step": 3649 }, { "epoch": 0.36, "learning_rate": 0.0014818264216684008, "loss": 1.6787, "step": 3650 }, { "epoch": 0.36, "learning_rate": 0.0014815466051637698, "loss": 1.6582, "step": 3651 }, { "epoch": 0.36, "learning_rate": 0.0014812667395642657, "loss": 1.6182, "step": 3652 }, { "epoch": 0.36, "learning_rate": 0.001480986824898421, "loss": 1.7119, "step": 3653 }, { "epoch": 0.36, "learning_rate": 0.0014807068611947733, "loss": 1.667, "step": 3654 }, { "epoch": 0.36, "learning_rate": 0.0014804268484818661, "loss": 1.6758, "step": 3655 }, { "epoch": 0.36, "learning_rate": 0.0014801467867882477, "loss": 1.6016, "step": 3656 }, { "epoch": 0.36, "learning_rate": 0.0014798666761424704, "loss": 1.8027, "step": 3657 }, { "epoch": 0.36, "learning_rate": 0.0014795865165730926, "loss": 1.624, "step": 3658 }, { "epoch": 0.36, "learning_rate": 0.0014793063081086771, "loss": 1.7148, "step": 3659 }, { "epoch": 0.36, "learning_rate": 0.0014790260507777915, "loss": 1.5752, "step": 3660 }, { "epoch": 0.36, "learning_rate": 0.0014787457446090093, "loss": 1.6582, "step": 3661 }, { "epoch": 0.36, "learning_rate": 0.0014784653896309085, "loss": 1.5576, "step": 3662 }, { "epoch": 0.36, "learning_rate": 0.0014781849858720715, "loss": 1.6016, "step": 3663 }, { "epoch": 0.36, "learning_rate": 0.0014779045333610864, "loss": 1.6533, "step": 3664 }, { "epoch": 0.36, "learning_rate": 0.001477624032126546, "loss": 1.6543, "step": 3665 }, { "epoch": 0.36, "learning_rate": 0.0014773434821970482, "loss": 1.709, "step": 3666 }, { "epoch": 0.36, "learning_rate": 0.0014770628836011954, "loss": 1.5977, "step": 3667 }, { "epoch": 0.36, "learning_rate": 0.0014767822363675957, "loss": 1.6299, "step": 3668 }, { "epoch": 0.36, "learning_rate": 0.0014765015405248617, "loss": 1.7246, "step": 3669 }, { "epoch": 0.36, "learning_rate": 0.001476220796101611, "loss": 1.6172, "step": 3670 }, { "epoch": 0.36, "learning_rate": 0.0014759400031264663, "loss": 1.6885, "step": 3671 }, { "epoch": 0.36, "learning_rate": 0.0014756591616280547, "loss": 1.6621, "step": 3672 }, { "epoch": 0.36, "learning_rate": 0.0014753782716350093, "loss": 1.6221, "step": 3673 }, { "epoch": 0.36, "learning_rate": 0.0014750973331759672, "loss": 1.5039, "step": 3674 }, { "epoch": 0.36, "learning_rate": 0.0014748163462795709, "loss": 1.7471, "step": 3675 }, { "epoch": 0.36, "learning_rate": 0.0014745353109744674, "loss": 1.6191, "step": 3676 }, { "epoch": 0.36, "learning_rate": 0.0014742542272893093, "loss": 1.6914, "step": 3677 }, { "epoch": 0.36, "learning_rate": 0.001473973095252754, "loss": 1.7031, "step": 3678 }, { "epoch": 0.36, "learning_rate": 0.001473691914893463, "loss": 1.6377, "step": 3679 }, { "epoch": 0.36, "learning_rate": 0.0014734106862401035, "loss": 1.7275, "step": 3680 }, { "epoch": 0.36, "learning_rate": 0.001473129409321348, "loss": 1.6357, "step": 3681 }, { "epoch": 0.36, "learning_rate": 0.0014728480841658725, "loss": 1.6396, "step": 3682 }, { "epoch": 0.36, "learning_rate": 0.0014725667108023595, "loss": 1.6338, "step": 3683 }, { "epoch": 0.36, "learning_rate": 0.0014722852892594953, "loss": 1.6689, "step": 3684 }, { "epoch": 0.36, "learning_rate": 0.0014720038195659717, "loss": 1.5752, "step": 3685 }, { "epoch": 0.36, "learning_rate": 0.0014717223017504855, "loss": 1.7217, "step": 3686 }, { "epoch": 0.36, "learning_rate": 0.0014714407358417377, "loss": 1.7461, "step": 3687 }, { "epoch": 0.36, "learning_rate": 0.0014711591218684351, "loss": 1.5996, "step": 3688 }, { "epoch": 0.36, "learning_rate": 0.0014708774598592882, "loss": 1.6299, "step": 3689 }, { "epoch": 0.36, "learning_rate": 0.0014705957498430137, "loss": 1.6309, "step": 3690 }, { "epoch": 0.36, "learning_rate": 0.0014703139918483329, "loss": 1.7969, "step": 3691 }, { "epoch": 0.36, "learning_rate": 0.001470032185903971, "loss": 1.7109, "step": 3692 }, { "epoch": 0.36, "learning_rate": 0.0014697503320386595, "loss": 1.7031, "step": 3693 }, { "epoch": 0.36, "learning_rate": 0.0014694684302811337, "loss": 1.4629, "step": 3694 }, { "epoch": 0.36, "learning_rate": 0.0014691864806601342, "loss": 1.5537, "step": 3695 }, { "epoch": 0.36, "learning_rate": 0.0014689044832044064, "loss": 1.6172, "step": 3696 }, { "epoch": 0.36, "learning_rate": 0.0014686224379427011, "loss": 1.668, "step": 3697 }, { "epoch": 0.36, "learning_rate": 0.0014683403449037731, "loss": 1.7139, "step": 3698 }, { "epoch": 0.36, "learning_rate": 0.0014680582041163826, "loss": 1.5303, "step": 3699 }, { "epoch": 0.36, "learning_rate": 0.0014677760156092945, "loss": 1.7148, "step": 3700 }, { "epoch": 0.36, "learning_rate": 0.0014674937794112786, "loss": 1.542, "step": 3701 }, { "epoch": 0.36, "learning_rate": 0.0014672114955511096, "loss": 1.7656, "step": 3702 }, { "epoch": 0.37, "learning_rate": 0.001466929164057567, "loss": 1.7812, "step": 3703 }, { "epoch": 0.37, "learning_rate": 0.0014666467849594356, "loss": 1.6338, "step": 3704 }, { "epoch": 0.37, "learning_rate": 0.001466364358285504, "loss": 1.6113, "step": 3705 }, { "epoch": 0.37, "learning_rate": 0.001466081884064567, "loss": 1.6885, "step": 3706 }, { "epoch": 0.37, "learning_rate": 0.001465799362325423, "loss": 1.583, "step": 3707 }, { "epoch": 0.37, "learning_rate": 0.0014655167930968758, "loss": 1.6533, "step": 3708 }, { "epoch": 0.37, "learning_rate": 0.0014652341764077343, "loss": 1.6592, "step": 3709 }, { "epoch": 0.37, "learning_rate": 0.0014649515122868123, "loss": 1.6592, "step": 3710 }, { "epoch": 0.37, "learning_rate": 0.0014646688007629272, "loss": 1.6172, "step": 3711 }, { "epoch": 0.37, "learning_rate": 0.0014643860418649026, "loss": 1.7363, "step": 3712 }, { "epoch": 0.37, "learning_rate": 0.001464103235621567, "loss": 1.707, "step": 3713 }, { "epoch": 0.37, "learning_rate": 0.0014638203820617523, "loss": 1.5039, "step": 3714 }, { "epoch": 0.37, "learning_rate": 0.0014635374812142965, "loss": 1.7744, "step": 3715 }, { "epoch": 0.37, "learning_rate": 0.0014632545331080423, "loss": 1.6807, "step": 3716 }, { "epoch": 0.37, "learning_rate": 0.0014629715377718368, "loss": 1.7012, "step": 3717 }, { "epoch": 0.37, "learning_rate": 0.001462688495234532, "loss": 1.6621, "step": 3718 }, { "epoch": 0.37, "learning_rate": 0.0014624054055249846, "loss": 1.6406, "step": 3719 }, { "epoch": 0.37, "learning_rate": 0.0014621222686720566, "loss": 1.7256, "step": 3720 }, { "epoch": 0.37, "learning_rate": 0.0014618390847046146, "loss": 1.6484, "step": 3721 }, { "epoch": 0.37, "learning_rate": 0.0014615558536515295, "loss": 1.5928, "step": 3722 }, { "epoch": 0.37, "learning_rate": 0.0014612725755416777, "loss": 1.6143, "step": 3723 }, { "epoch": 0.37, "learning_rate": 0.0014609892504039402, "loss": 1.7129, "step": 3724 }, { "epoch": 0.37, "learning_rate": 0.0014607058782672023, "loss": 1.6387, "step": 3725 }, { "epoch": 0.37, "learning_rate": 0.0014604224591603548, "loss": 1.6357, "step": 3726 }, { "epoch": 0.37, "learning_rate": 0.0014601389931122929, "loss": 1.6934, "step": 3727 }, { "epoch": 0.37, "learning_rate": 0.0014598554801519166, "loss": 1.7393, "step": 3728 }, { "epoch": 0.37, "learning_rate": 0.001459571920308131, "loss": 1.6738, "step": 3729 }, { "epoch": 0.37, "learning_rate": 0.0014592883136098455, "loss": 1.709, "step": 3730 }, { "epoch": 0.37, "learning_rate": 0.0014590046600859747, "loss": 1.6934, "step": 3731 }, { "epoch": 0.37, "learning_rate": 0.0014587209597654379, "loss": 1.6436, "step": 3732 }, { "epoch": 0.37, "learning_rate": 0.0014584372126771584, "loss": 1.6406, "step": 3733 }, { "epoch": 0.37, "learning_rate": 0.0014581534188500653, "loss": 1.6426, "step": 3734 }, { "epoch": 0.37, "learning_rate": 0.0014578695783130924, "loss": 1.7236, "step": 3735 }, { "epoch": 0.37, "learning_rate": 0.0014575856910951776, "loss": 1.71, "step": 3736 }, { "epoch": 0.37, "learning_rate": 0.0014573017572252639, "loss": 1.6826, "step": 3737 }, { "epoch": 0.37, "learning_rate": 0.001457017776732299, "loss": 1.5635, "step": 3738 }, { "epoch": 0.37, "learning_rate": 0.0014567337496452358, "loss": 1.5889, "step": 3739 }, { "epoch": 0.37, "learning_rate": 0.0014564496759930312, "loss": 1.6631, "step": 3740 }, { "epoch": 0.37, "learning_rate": 0.0014561655558046477, "loss": 1.6328, "step": 3741 }, { "epoch": 0.37, "learning_rate": 0.0014558813891090512, "loss": 1.5859, "step": 3742 }, { "epoch": 0.37, "learning_rate": 0.001455597175935214, "loss": 1.7305, "step": 3743 }, { "epoch": 0.37, "learning_rate": 0.001455312916312112, "loss": 1.6094, "step": 3744 }, { "epoch": 0.37, "learning_rate": 0.0014550286102687259, "loss": 1.5811, "step": 3745 }, { "epoch": 0.37, "learning_rate": 0.0014547442578340419, "loss": 1.5908, "step": 3746 }, { "epoch": 0.37, "learning_rate": 0.0014544598590370504, "loss": 1.5225, "step": 3747 }, { "epoch": 0.37, "learning_rate": 0.0014541754139067464, "loss": 1.5889, "step": 3748 }, { "epoch": 0.37, "learning_rate": 0.0014538909224721301, "loss": 1.5625, "step": 3749 }, { "epoch": 0.37, "learning_rate": 0.0014536063847622057, "loss": 1.6924, "step": 3750 }, { "epoch": 0.37, "learning_rate": 0.0014533218008059823, "loss": 1.7646, "step": 3751 }, { "epoch": 0.37, "learning_rate": 0.0014530371706324744, "loss": 1.4834, "step": 3752 }, { "epoch": 0.37, "learning_rate": 0.0014527524942707006, "loss": 1.6895, "step": 3753 }, { "epoch": 0.37, "learning_rate": 0.0014524677717496846, "loss": 1.5986, "step": 3754 }, { "epoch": 0.37, "learning_rate": 0.0014521830030984543, "loss": 1.6846, "step": 3755 }, { "epoch": 0.37, "learning_rate": 0.0014518981883460426, "loss": 1.7041, "step": 3756 }, { "epoch": 0.37, "learning_rate": 0.0014516133275214874, "loss": 1.666, "step": 3757 }, { "epoch": 0.37, "learning_rate": 0.0014513284206538306, "loss": 1.7168, "step": 3758 }, { "epoch": 0.37, "learning_rate": 0.0014510434677721192, "loss": 1.7207, "step": 3759 }, { "epoch": 0.37, "learning_rate": 0.0014507584689054046, "loss": 1.6611, "step": 3760 }, { "epoch": 0.37, "learning_rate": 0.0014504734240827438, "loss": 1.5332, "step": 3761 }, { "epoch": 0.37, "learning_rate": 0.0014501883333331972, "loss": 1.5186, "step": 3762 }, { "epoch": 0.37, "learning_rate": 0.0014499031966858312, "loss": 1.6494, "step": 3763 }, { "epoch": 0.37, "learning_rate": 0.0014496180141697155, "loss": 1.7402, "step": 3764 }, { "epoch": 0.37, "learning_rate": 0.0014493327858139257, "loss": 1.5752, "step": 3765 }, { "epoch": 0.37, "learning_rate": 0.001449047511647541, "loss": 1.6387, "step": 3766 }, { "epoch": 0.37, "learning_rate": 0.0014487621916996463, "loss": 1.666, "step": 3767 }, { "epoch": 0.37, "learning_rate": 0.0014484768259993303, "loss": 1.6475, "step": 3768 }, { "epoch": 0.37, "learning_rate": 0.001448191414575687, "loss": 1.6299, "step": 3769 }, { "epoch": 0.37, "learning_rate": 0.0014479059574578146, "loss": 1.6143, "step": 3770 }, { "epoch": 0.37, "learning_rate": 0.0014476204546748166, "loss": 1.5762, "step": 3771 }, { "epoch": 0.37, "learning_rate": 0.0014473349062558002, "loss": 1.6719, "step": 3772 }, { "epoch": 0.37, "learning_rate": 0.001447049312229878, "loss": 1.6484, "step": 3773 }, { "epoch": 0.37, "learning_rate": 0.001446763672626167, "loss": 1.5146, "step": 3774 }, { "epoch": 0.37, "learning_rate": 0.001446477987473789, "loss": 1.7002, "step": 3775 }, { "epoch": 0.37, "learning_rate": 0.00144619225680187, "loss": 1.5684, "step": 3776 }, { "epoch": 0.37, "learning_rate": 0.0014459064806395412, "loss": 1.5596, "step": 3777 }, { "epoch": 0.37, "learning_rate": 0.0014456206590159383, "loss": 1.6445, "step": 3778 }, { "epoch": 0.37, "learning_rate": 0.0014453347919602011, "loss": 1.6445, "step": 3779 }, { "epoch": 0.37, "learning_rate": 0.0014450488795014747, "loss": 1.5781, "step": 3780 }, { "epoch": 0.37, "learning_rate": 0.0014447629216689086, "loss": 1.7168, "step": 3781 }, { "epoch": 0.37, "learning_rate": 0.0014444769184916572, "loss": 1.5635, "step": 3782 }, { "epoch": 0.37, "learning_rate": 0.0014441908699988786, "loss": 1.6133, "step": 3783 }, { "epoch": 0.37, "learning_rate": 0.0014439047762197367, "loss": 1.6094, "step": 3784 }, { "epoch": 0.37, "learning_rate": 0.001443618637183399, "loss": 1.5391, "step": 3785 }, { "epoch": 0.37, "learning_rate": 0.0014433324529190383, "loss": 1.6855, "step": 3786 }, { "epoch": 0.37, "learning_rate": 0.001443046223455832, "loss": 1.6094, "step": 3787 }, { "epoch": 0.37, "learning_rate": 0.0014427599488229619, "loss": 1.6201, "step": 3788 }, { "epoch": 0.37, "learning_rate": 0.0014424736290496144, "loss": 1.7109, "step": 3789 }, { "epoch": 0.37, "learning_rate": 0.00144218726416498, "loss": 1.7705, "step": 3790 }, { "epoch": 0.37, "learning_rate": 0.0014419008541982547, "loss": 1.5996, "step": 3791 }, { "epoch": 0.37, "learning_rate": 0.001441614399178639, "loss": 1.7295, "step": 3792 }, { "epoch": 0.37, "learning_rate": 0.0014413278991353372, "loss": 1.5264, "step": 3793 }, { "epoch": 0.37, "learning_rate": 0.001441041354097559, "loss": 1.5898, "step": 3794 }, { "epoch": 0.37, "learning_rate": 0.0014407547640945177, "loss": 1.6172, "step": 3795 }, { "epoch": 0.37, "learning_rate": 0.0014404681291554332, "loss": 1.4883, "step": 3796 }, { "epoch": 0.37, "learning_rate": 0.0014401814493095273, "loss": 1.6621, "step": 3797 }, { "epoch": 0.37, "learning_rate": 0.0014398947245860287, "loss": 1.6211, "step": 3798 }, { "epoch": 0.37, "learning_rate": 0.0014396079550141692, "loss": 1.5859, "step": 3799 }, { "epoch": 0.37, "learning_rate": 0.0014393211406231855, "loss": 1.7188, "step": 3800 }, { "epoch": 0.37, "learning_rate": 0.0014390342814423198, "loss": 1.6152, "step": 3801 }, { "epoch": 0.37, "learning_rate": 0.0014387473775008175, "loss": 1.748, "step": 3802 }, { "epoch": 0.37, "learning_rate": 0.001438460428827929, "loss": 1.6191, "step": 3803 }, { "epoch": 0.38, "learning_rate": 0.0014381734354529097, "loss": 1.5791, "step": 3804 }, { "epoch": 0.38, "learning_rate": 0.00143788639740502, "loss": 1.7939, "step": 3805 }, { "epoch": 0.38, "learning_rate": 0.0014375993147135227, "loss": 1.5059, "step": 3806 }, { "epoch": 0.38, "learning_rate": 0.0014373121874076875, "loss": 1.6699, "step": 3807 }, { "epoch": 0.38, "learning_rate": 0.0014370250155167877, "loss": 1.6807, "step": 3808 }, { "epoch": 0.38, "learning_rate": 0.0014367377990701009, "loss": 1.5928, "step": 3809 }, { "epoch": 0.38, "learning_rate": 0.00143645053809691, "loss": 1.6807, "step": 3810 }, { "epoch": 0.38, "learning_rate": 0.0014361632326265012, "loss": 1.6338, "step": 3811 }, { "epoch": 0.38, "learning_rate": 0.0014358758826881668, "loss": 1.6162, "step": 3812 }, { "epoch": 0.38, "learning_rate": 0.0014355884883112025, "loss": 1.749, "step": 3813 }, { "epoch": 0.38, "learning_rate": 0.0014353010495249089, "loss": 1.5234, "step": 3814 }, { "epoch": 0.38, "learning_rate": 0.001435013566358591, "loss": 1.5938, "step": 3815 }, { "epoch": 0.38, "learning_rate": 0.0014347260388415585, "loss": 1.5137, "step": 3816 }, { "epoch": 0.38, "learning_rate": 0.0014344384670031257, "loss": 1.7354, "step": 3817 }, { "epoch": 0.38, "learning_rate": 0.001434150850872611, "loss": 1.584, "step": 3818 }, { "epoch": 0.38, "learning_rate": 0.0014338631904793379, "loss": 1.5693, "step": 3819 }, { "epoch": 0.38, "learning_rate": 0.0014335754858526336, "loss": 1.6641, "step": 3820 }, { "epoch": 0.38, "learning_rate": 0.0014332877370218306, "loss": 1.5088, "step": 3821 }, { "epoch": 0.38, "learning_rate": 0.0014329999440162657, "loss": 1.6182, "step": 3822 }, { "epoch": 0.38, "learning_rate": 0.0014327121068652801, "loss": 1.6953, "step": 3823 }, { "epoch": 0.38, "learning_rate": 0.0014324242255982194, "loss": 1.6016, "step": 3824 }, { "epoch": 0.38, "learning_rate": 0.0014321363002444338, "loss": 1.6582, "step": 3825 }, { "epoch": 0.38, "learning_rate": 0.0014318483308332782, "loss": 1.5029, "step": 3826 }, { "epoch": 0.38, "learning_rate": 0.0014315603173941116, "loss": 1.5762, "step": 3827 }, { "epoch": 0.38, "learning_rate": 0.0014312722599562975, "loss": 1.5928, "step": 3828 }, { "epoch": 0.38, "learning_rate": 0.0014309841585492043, "loss": 1.6836, "step": 3829 }, { "epoch": 0.38, "learning_rate": 0.001430696013202205, "loss": 1.5547, "step": 3830 }, { "epoch": 0.38, "learning_rate": 0.001430407823944676, "loss": 1.6992, "step": 3831 }, { "epoch": 0.38, "learning_rate": 0.0014301195908059997, "loss": 1.5098, "step": 3832 }, { "epoch": 0.38, "learning_rate": 0.0014298313138155615, "loss": 1.6973, "step": 3833 }, { "epoch": 0.38, "learning_rate": 0.0014295429930027525, "loss": 1.627, "step": 3834 }, { "epoch": 0.38, "learning_rate": 0.0014292546283969673, "loss": 1.6045, "step": 3835 }, { "epoch": 0.38, "learning_rate": 0.0014289662200276056, "loss": 1.6426, "step": 3836 }, { "epoch": 0.38, "learning_rate": 0.0014286777679240713, "loss": 1.6357, "step": 3837 }, { "epoch": 0.38, "learning_rate": 0.0014283892721157728, "loss": 1.6768, "step": 3838 }, { "epoch": 0.38, "learning_rate": 0.0014281007326321228, "loss": 1.6211, "step": 3839 }, { "epoch": 0.38, "learning_rate": 0.0014278121495025391, "loss": 1.75, "step": 3840 }, { "epoch": 0.38, "learning_rate": 0.001427523522756443, "loss": 1.5771, "step": 3841 }, { "epoch": 0.38, "learning_rate": 0.0014272348524232608, "loss": 1.6191, "step": 3842 }, { "epoch": 0.38, "learning_rate": 0.0014269461385324233, "loss": 1.6514, "step": 3843 }, { "epoch": 0.38, "learning_rate": 0.0014266573811133657, "loss": 1.5752, "step": 3844 }, { "epoch": 0.38, "learning_rate": 0.0014263685801955272, "loss": 1.5801, "step": 3845 }, { "epoch": 0.38, "learning_rate": 0.001426079735808352, "loss": 1.6055, "step": 3846 }, { "epoch": 0.38, "learning_rate": 0.0014257908479812884, "loss": 1.5566, "step": 3847 }, { "epoch": 0.38, "learning_rate": 0.0014255019167437892, "loss": 1.6201, "step": 3848 }, { "epoch": 0.38, "learning_rate": 0.001425212942125312, "loss": 1.7285, "step": 3849 }, { "epoch": 0.38, "learning_rate": 0.001424923924155318, "loss": 1.6865, "step": 3850 }, { "epoch": 0.38, "learning_rate": 0.0014246348628632738, "loss": 1.6846, "step": 3851 }, { "epoch": 0.38, "learning_rate": 0.0014243457582786493, "loss": 1.625, "step": 3852 }, { "epoch": 0.38, "learning_rate": 0.00142405661043092, "loss": 1.5635, "step": 3853 }, { "epoch": 0.38, "learning_rate": 0.0014237674193495647, "loss": 1.5771, "step": 3854 }, { "epoch": 0.38, "learning_rate": 0.0014234781850640677, "loss": 1.5625, "step": 3855 }, { "epoch": 0.38, "learning_rate": 0.0014231889076039169, "loss": 1.5586, "step": 3856 }, { "epoch": 0.38, "learning_rate": 0.0014228995869986047, "loss": 1.5898, "step": 3857 }, { "epoch": 0.38, "learning_rate": 0.0014226102232776285, "loss": 1.6865, "step": 3858 }, { "epoch": 0.38, "learning_rate": 0.0014223208164704892, "loss": 1.6514, "step": 3859 }, { "epoch": 0.38, "learning_rate": 0.001422031366606693, "loss": 1.5586, "step": 3860 }, { "epoch": 0.38, "learning_rate": 0.0014217418737157495, "loss": 1.6807, "step": 3861 }, { "epoch": 0.38, "learning_rate": 0.0014214523378271735, "loss": 1.7129, "step": 3862 }, { "epoch": 0.38, "learning_rate": 0.0014211627589704838, "loss": 1.5967, "step": 3863 }, { "epoch": 0.38, "learning_rate": 0.001420873137175204, "loss": 1.5859, "step": 3864 }, { "epoch": 0.38, "learning_rate": 0.0014205834724708613, "loss": 1.6934, "step": 3865 }, { "epoch": 0.38, "learning_rate": 0.001420293764886988, "loss": 1.5381, "step": 3866 }, { "epoch": 0.38, "learning_rate": 0.0014200040144531207, "loss": 1.5566, "step": 3867 }, { "epoch": 0.38, "learning_rate": 0.0014197142211987996, "loss": 1.6553, "step": 3868 }, { "epoch": 0.38, "learning_rate": 0.0014194243851535704, "loss": 1.5137, "step": 3869 }, { "epoch": 0.38, "learning_rate": 0.0014191345063469822, "loss": 1.6133, "step": 3870 }, { "epoch": 0.38, "learning_rate": 0.0014188445848085892, "loss": 1.7041, "step": 3871 }, { "epoch": 0.38, "learning_rate": 0.0014185546205679493, "loss": 1.6582, "step": 3872 }, { "epoch": 0.38, "learning_rate": 0.0014182646136546256, "loss": 1.6309, "step": 3873 }, { "epoch": 0.38, "learning_rate": 0.0014179745640981844, "loss": 1.5537, "step": 3874 }, { "epoch": 0.38, "learning_rate": 0.0014176844719281975, "loss": 1.5645, "step": 3875 }, { "epoch": 0.38, "learning_rate": 0.0014173943371742401, "loss": 1.7061, "step": 3876 }, { "epoch": 0.38, "learning_rate": 0.0014171041598658924, "loss": 1.7217, "step": 3877 }, { "epoch": 0.38, "learning_rate": 0.0014168139400327388, "loss": 1.5908, "step": 3878 }, { "epoch": 0.38, "learning_rate": 0.001416523677704367, "loss": 1.627, "step": 3879 }, { "epoch": 0.38, "learning_rate": 0.0014162333729103717, "loss": 1.6396, "step": 3880 }, { "epoch": 0.38, "learning_rate": 0.0014159430256803485, "loss": 1.5898, "step": 3881 }, { "epoch": 0.38, "learning_rate": 0.0014156526360439002, "loss": 1.665, "step": 3882 }, { "epoch": 0.38, "learning_rate": 0.001415362204030632, "loss": 1.6543, "step": 3883 }, { "epoch": 0.38, "learning_rate": 0.0014150717296701548, "loss": 1.6553, "step": 3884 }, { "epoch": 0.38, "learning_rate": 0.0014147812129920827, "loss": 1.6494, "step": 3885 }, { "epoch": 0.38, "learning_rate": 0.0014144906540260347, "loss": 1.6826, "step": 3886 }, { "epoch": 0.38, "learning_rate": 0.001414200052801634, "loss": 1.7354, "step": 3887 }, { "epoch": 0.38, "learning_rate": 0.0014139094093485079, "loss": 1.5615, "step": 3888 }, { "epoch": 0.38, "learning_rate": 0.001413618723696289, "loss": 1.5986, "step": 3889 }, { "epoch": 0.38, "learning_rate": 0.0014133279958746128, "loss": 1.5615, "step": 3890 }, { "epoch": 0.38, "learning_rate": 0.0014130372259131198, "loss": 1.5908, "step": 3891 }, { "epoch": 0.38, "learning_rate": 0.0014127464138414547, "loss": 1.623, "step": 3892 }, { "epoch": 0.38, "learning_rate": 0.0014124555596892665, "loss": 1.6396, "step": 3893 }, { "epoch": 0.38, "learning_rate": 0.001412164663486209, "loss": 1.6348, "step": 3894 }, { "epoch": 0.38, "learning_rate": 0.0014118737252619385, "loss": 1.6699, "step": 3895 }, { "epoch": 0.38, "learning_rate": 0.0014115827450461183, "loss": 1.6074, "step": 3896 }, { "epoch": 0.38, "learning_rate": 0.0014112917228684138, "loss": 1.4844, "step": 3897 }, { "epoch": 0.38, "learning_rate": 0.001411000658758496, "loss": 1.6572, "step": 3898 }, { "epoch": 0.38, "learning_rate": 0.001410709552746039, "loss": 1.5869, "step": 3899 }, { "epoch": 0.38, "learning_rate": 0.001410418404860722, "loss": 1.6973, "step": 3900 }, { "epoch": 0.38, "learning_rate": 0.0014101272151322284, "loss": 1.54, "step": 3901 }, { "epoch": 0.38, "learning_rate": 0.0014098359835902454, "loss": 1.6992, "step": 3902 }, { "epoch": 0.38, "learning_rate": 0.0014095447102644653, "loss": 1.8125, "step": 3903 }, { "epoch": 0.38, "learning_rate": 0.0014092533951845835, "loss": 1.5781, "step": 3904 }, { "epoch": 0.38, "learning_rate": 0.0014089620383803006, "loss": 1.8096, "step": 3905 }, { "epoch": 0.39, "learning_rate": 0.0014086706398813212, "loss": 1.5928, "step": 3906 }, { "epoch": 0.39, "learning_rate": 0.0014083791997173544, "loss": 1.7119, "step": 3907 }, { "epoch": 0.39, "learning_rate": 0.0014080877179181125, "loss": 1.7002, "step": 3908 }, { "epoch": 0.39, "learning_rate": 0.0014077961945133132, "loss": 1.7354, "step": 3909 }, { "epoch": 0.39, "learning_rate": 0.0014075046295326783, "loss": 1.5869, "step": 3910 }, { "epoch": 0.39, "learning_rate": 0.0014072130230059331, "loss": 1.6279, "step": 3911 }, { "epoch": 0.39, "learning_rate": 0.001406921374962808, "loss": 1.6895, "step": 3912 }, { "epoch": 0.39, "learning_rate": 0.0014066296854330367, "loss": 1.6582, "step": 3913 }, { "epoch": 0.39, "learning_rate": 0.0014063379544463584, "loss": 1.4541, "step": 3914 }, { "epoch": 0.39, "learning_rate": 0.0014060461820325151, "loss": 1.6123, "step": 3915 }, { "epoch": 0.39, "learning_rate": 0.0014057543682212546, "loss": 1.6113, "step": 3916 }, { "epoch": 0.39, "learning_rate": 0.001405462513042327, "loss": 1.5977, "step": 3917 }, { "epoch": 0.39, "learning_rate": 0.0014051706165254887, "loss": 1.5107, "step": 3918 }, { "epoch": 0.39, "learning_rate": 0.0014048786787004986, "loss": 1.624, "step": 3919 }, { "epoch": 0.39, "learning_rate": 0.0014045866995971205, "loss": 1.6895, "step": 3920 }, { "epoch": 0.39, "learning_rate": 0.0014042946792451221, "loss": 1.5674, "step": 3921 }, { "epoch": 0.39, "learning_rate": 0.0014040026176742764, "loss": 1.668, "step": 3922 }, { "epoch": 0.39, "learning_rate": 0.0014037105149143595, "loss": 1.5771, "step": 3923 }, { "epoch": 0.39, "learning_rate": 0.0014034183709951517, "loss": 1.5254, "step": 3924 }, { "epoch": 0.39, "learning_rate": 0.0014031261859464384, "loss": 1.75, "step": 3925 }, { "epoch": 0.39, "learning_rate": 0.001402833959798008, "loss": 1.6279, "step": 3926 }, { "epoch": 0.39, "learning_rate": 0.001402541692579654, "loss": 1.6279, "step": 3927 }, { "epoch": 0.39, "learning_rate": 0.0014022493843211734, "loss": 1.626, "step": 3928 }, { "epoch": 0.39, "learning_rate": 0.0014019570350523685, "loss": 1.6504, "step": 3929 }, { "epoch": 0.39, "learning_rate": 0.001401664644803044, "loss": 1.6357, "step": 3930 }, { "epoch": 0.39, "learning_rate": 0.0014013722136030106, "loss": 1.6387, "step": 3931 }, { "epoch": 0.39, "learning_rate": 0.001401079741482082, "loss": 1.7021, "step": 3932 }, { "epoch": 0.39, "learning_rate": 0.0014007872284700772, "loss": 1.7539, "step": 3933 }, { "epoch": 0.39, "learning_rate": 0.0014004946745968176, "loss": 1.543, "step": 3934 }, { "epoch": 0.39, "learning_rate": 0.0014002020798921304, "loss": 1.625, "step": 3935 }, { "epoch": 0.39, "learning_rate": 0.0013999094443858462, "loss": 1.6465, "step": 3936 }, { "epoch": 0.39, "learning_rate": 0.0013996167681077998, "loss": 1.4902, "step": 3937 }, { "epoch": 0.39, "learning_rate": 0.0013993240510878305, "loss": 1.6318, "step": 3938 }, { "epoch": 0.39, "learning_rate": 0.001399031293355781, "loss": 1.583, "step": 3939 }, { "epoch": 0.39, "learning_rate": 0.0013987384949414995, "loss": 1.7471, "step": 3940 }, { "epoch": 0.39, "learning_rate": 0.0013984456558748372, "loss": 1.6162, "step": 3941 }, { "epoch": 0.39, "learning_rate": 0.0013981527761856498, "loss": 1.6416, "step": 3942 }, { "epoch": 0.39, "learning_rate": 0.0013978598559037966, "loss": 1.6035, "step": 3943 }, { "epoch": 0.39, "learning_rate": 0.0013975668950591425, "loss": 1.6172, "step": 3944 }, { "epoch": 0.39, "learning_rate": 0.0013972738936815548, "loss": 1.6045, "step": 3945 }, { "epoch": 0.39, "learning_rate": 0.001396980851800906, "loss": 1.5684, "step": 3946 }, { "epoch": 0.39, "learning_rate": 0.001396687769447072, "loss": 1.6465, "step": 3947 }, { "epoch": 0.39, "learning_rate": 0.0013963946466499341, "loss": 1.585, "step": 3948 }, { "epoch": 0.39, "learning_rate": 0.0013961014834393763, "loss": 1.7002, "step": 3949 }, { "epoch": 0.39, "learning_rate": 0.001395808279845288, "loss": 1.5605, "step": 3950 }, { "epoch": 0.39, "learning_rate": 0.001395515035897561, "loss": 1.6074, "step": 3951 }, { "epoch": 0.39, "learning_rate": 0.001395221751626093, "loss": 1.6455, "step": 3952 }, { "epoch": 0.39, "learning_rate": 0.0013949284270607847, "loss": 1.8174, "step": 3953 }, { "epoch": 0.39, "learning_rate": 0.0013946350622315415, "loss": 1.6367, "step": 3954 }, { "epoch": 0.39, "learning_rate": 0.0013943416571682726, "loss": 1.6133, "step": 3955 }, { "epoch": 0.39, "learning_rate": 0.0013940482119008913, "loss": 1.7188, "step": 3956 }, { "epoch": 0.39, "learning_rate": 0.0013937547264593153, "loss": 1.6816, "step": 3957 }, { "epoch": 0.39, "learning_rate": 0.001393461200873466, "loss": 1.5928, "step": 3958 }, { "epoch": 0.39, "learning_rate": 0.0013931676351732689, "loss": 1.5781, "step": 3959 }, { "epoch": 0.39, "learning_rate": 0.0013928740293886542, "loss": 1.5537, "step": 3960 }, { "epoch": 0.39, "learning_rate": 0.0013925803835495552, "loss": 1.5615, "step": 3961 }, { "epoch": 0.39, "learning_rate": 0.0013922866976859103, "loss": 1.5693, "step": 3962 }, { "epoch": 0.39, "learning_rate": 0.0013919929718276612, "loss": 1.5986, "step": 3963 }, { "epoch": 0.39, "learning_rate": 0.001391699206004754, "loss": 1.5303, "step": 3964 }, { "epoch": 0.39, "learning_rate": 0.001391405400247139, "loss": 1.6533, "step": 3965 }, { "epoch": 0.39, "learning_rate": 0.0013911115545847702, "loss": 1.6963, "step": 3966 }, { "epoch": 0.39, "learning_rate": 0.0013908176690476065, "loss": 1.6523, "step": 3967 }, { "epoch": 0.39, "learning_rate": 0.0013905237436656096, "loss": 1.6914, "step": 3968 }, { "epoch": 0.39, "learning_rate": 0.001390229778468746, "loss": 1.6797, "step": 3969 }, { "epoch": 0.39, "learning_rate": 0.0013899357734869864, "loss": 1.5527, "step": 3970 }, { "epoch": 0.39, "learning_rate": 0.0013896417287503054, "loss": 1.6416, "step": 3971 }, { "epoch": 0.39, "learning_rate": 0.0013893476442886814, "loss": 1.4785, "step": 3972 }, { "epoch": 0.39, "learning_rate": 0.0013890535201320968, "loss": 1.6895, "step": 3973 }, { "epoch": 0.39, "learning_rate": 0.0013887593563105393, "loss": 1.6152, "step": 3974 }, { "epoch": 0.39, "learning_rate": 0.0013884651528539985, "loss": 1.582, "step": 3975 }, { "epoch": 0.39, "learning_rate": 0.0013881709097924694, "loss": 1.6191, "step": 3976 }, { "epoch": 0.39, "learning_rate": 0.0013878766271559512, "loss": 1.7725, "step": 3977 }, { "epoch": 0.39, "learning_rate": 0.0013875823049744468, "loss": 1.626, "step": 3978 }, { "epoch": 0.39, "learning_rate": 0.0013872879432779623, "loss": 1.4111, "step": 3979 }, { "epoch": 0.39, "learning_rate": 0.0013869935420965094, "loss": 1.6709, "step": 3980 }, { "epoch": 0.39, "learning_rate": 0.0013866991014601026, "loss": 1.5908, "step": 3981 }, { "epoch": 0.39, "learning_rate": 0.0013864046213987614, "loss": 1.5996, "step": 3982 }, { "epoch": 0.39, "learning_rate": 0.0013861101019425081, "loss": 1.583, "step": 3983 }, { "epoch": 0.39, "learning_rate": 0.0013858155431213702, "loss": 1.6436, "step": 3984 }, { "epoch": 0.39, "learning_rate": 0.0013855209449653782, "loss": 1.5781, "step": 3985 }, { "epoch": 0.39, "learning_rate": 0.0013852263075045676, "loss": 1.6895, "step": 3986 }, { "epoch": 0.39, "learning_rate": 0.0013849316307689775, "loss": 1.5938, "step": 3987 }, { "epoch": 0.39, "learning_rate": 0.00138463691478865, "loss": 1.7021, "step": 3988 }, { "epoch": 0.39, "learning_rate": 0.0013843421595936332, "loss": 1.6523, "step": 3989 }, { "epoch": 0.39, "learning_rate": 0.0013840473652139776, "loss": 1.6328, "step": 3990 }, { "epoch": 0.39, "learning_rate": 0.0013837525316797385, "loss": 1.5879, "step": 3991 }, { "epoch": 0.39, "learning_rate": 0.0013834576590209746, "loss": 1.6162, "step": 3992 }, { "epoch": 0.39, "learning_rate": 0.001383162747267749, "loss": 1.5361, "step": 3993 }, { "epoch": 0.39, "learning_rate": 0.001382867796450129, "loss": 1.624, "step": 3994 }, { "epoch": 0.39, "learning_rate": 0.001382572806598185, "loss": 1.6484, "step": 3995 }, { "epoch": 0.39, "learning_rate": 0.0013822777777419925, "loss": 1.6953, "step": 3996 }, { "epoch": 0.39, "learning_rate": 0.00138198270991163, "loss": 1.6055, "step": 3997 }, { "epoch": 0.39, "learning_rate": 0.0013816876031371803, "loss": 1.4873, "step": 3998 }, { "epoch": 0.39, "learning_rate": 0.001381392457448731, "loss": 1.6172, "step": 3999 }, { "epoch": 0.39, "learning_rate": 0.0013810972728763723, "loss": 1.584, "step": 4000 }, { "epoch": 0.39, "learning_rate": 0.0013808020494501991, "loss": 1.6611, "step": 4001 }, { "epoch": 0.39, "learning_rate": 0.0013805067872003105, "loss": 1.585, "step": 4002 }, { "epoch": 0.39, "learning_rate": 0.0013802114861568087, "loss": 1.6748, "step": 4003 }, { "epoch": 0.39, "learning_rate": 0.001379916146349801, "loss": 1.7119, "step": 4004 }, { "epoch": 0.39, "learning_rate": 0.001379620767809397, "loss": 1.5742, "step": 4005 }, { "epoch": 0.39, "learning_rate": 0.0013793253505657121, "loss": 1.6006, "step": 4006 }, { "epoch": 0.4, "learning_rate": 0.0013790298946488651, "loss": 1.7305, "step": 4007 }, { "epoch": 0.4, "learning_rate": 0.0013787344000889777, "loss": 1.5938, "step": 4008 }, { "epoch": 0.4, "learning_rate": 0.0013784388669161767, "loss": 1.4736, "step": 4009 }, { "epoch": 0.4, "learning_rate": 0.0013781432951605927, "loss": 1.6172, "step": 4010 }, { "epoch": 0.4, "learning_rate": 0.001377847684852359, "loss": 1.6445, "step": 4011 }, { "epoch": 0.4, "learning_rate": 0.001377552036021615, "loss": 1.5391, "step": 4012 }, { "epoch": 0.4, "learning_rate": 0.0013772563486985022, "loss": 1.6133, "step": 4013 }, { "epoch": 0.4, "learning_rate": 0.0013769606229131664, "loss": 1.5342, "step": 4014 }, { "epoch": 0.4, "learning_rate": 0.001376664858695758, "loss": 1.6699, "step": 4015 }, { "epoch": 0.4, "learning_rate": 0.001376369056076431, "loss": 1.6162, "step": 4016 }, { "epoch": 0.4, "learning_rate": 0.0013760732150853432, "loss": 1.7061, "step": 4017 }, { "epoch": 0.4, "learning_rate": 0.001375777335752656, "loss": 1.5615, "step": 4018 }, { "epoch": 0.4, "learning_rate": 0.001375481418108535, "loss": 1.6865, "step": 4019 }, { "epoch": 0.4, "learning_rate": 0.0013751854621831502, "loss": 1.6416, "step": 4020 }, { "epoch": 0.4, "learning_rate": 0.0013748894680066747, "loss": 1.8027, "step": 4021 }, { "epoch": 0.4, "learning_rate": 0.001374593435609286, "loss": 1.6758, "step": 4022 }, { "epoch": 0.4, "learning_rate": 0.0013742973650211648, "loss": 1.7842, "step": 4023 }, { "epoch": 0.4, "learning_rate": 0.001374001256272497, "loss": 1.5703, "step": 4024 }, { "epoch": 0.4, "learning_rate": 0.0013737051093934713, "loss": 1.5898, "step": 4025 }, { "epoch": 0.4, "learning_rate": 0.0013734089244142806, "loss": 1.5225, "step": 4026 }, { "epoch": 0.4, "learning_rate": 0.001373112701365122, "loss": 1.7051, "step": 4027 }, { "epoch": 0.4, "learning_rate": 0.0013728164402761955, "loss": 1.5293, "step": 4028 }, { "epoch": 0.4, "learning_rate": 0.0013725201411777061, "loss": 1.5811, "step": 4029 }, { "epoch": 0.4, "learning_rate": 0.0013722238040998625, "loss": 1.6523, "step": 4030 }, { "epoch": 0.4, "learning_rate": 0.0013719274290728764, "loss": 1.6855, "step": 4031 }, { "epoch": 0.4, "learning_rate": 0.0013716310161269645, "loss": 1.7031, "step": 4032 }, { "epoch": 0.4, "learning_rate": 0.0013713345652923464, "loss": 1.5762, "step": 4033 }, { "epoch": 0.4, "learning_rate": 0.001371038076599246, "loss": 1.7197, "step": 4034 }, { "epoch": 0.4, "learning_rate": 0.0013707415500778914, "loss": 1.627, "step": 4035 }, { "epoch": 0.4, "learning_rate": 0.0013704449857585145, "loss": 1.5693, "step": 4036 }, { "epoch": 0.4, "learning_rate": 0.0013701483836713498, "loss": 1.5273, "step": 4037 }, { "epoch": 0.4, "learning_rate": 0.0013698517438466371, "loss": 1.6133, "step": 4038 }, { "epoch": 0.4, "learning_rate": 0.0013695550663146198, "loss": 1.7344, "step": 4039 }, { "epoch": 0.4, "learning_rate": 0.0013692583511055445, "loss": 1.6738, "step": 4040 }, { "epoch": 0.4, "learning_rate": 0.0013689615982496626, "loss": 1.5205, "step": 4041 }, { "epoch": 0.4, "learning_rate": 0.0013686648077772285, "loss": 1.4844, "step": 4042 }, { "epoch": 0.4, "learning_rate": 0.0013683679797185004, "loss": 1.5732, "step": 4043 }, { "epoch": 0.4, "learning_rate": 0.0013680711141037412, "loss": 1.5713, "step": 4044 }, { "epoch": 0.4, "learning_rate": 0.001367774210963217, "loss": 1.5986, "step": 4045 }, { "epoch": 0.4, "learning_rate": 0.0013674772703271974, "loss": 1.6572, "step": 4046 }, { "epoch": 0.4, "learning_rate": 0.0013671802922259567, "loss": 1.6982, "step": 4047 }, { "epoch": 0.4, "learning_rate": 0.001366883276689772, "loss": 1.6533, "step": 4048 }, { "epoch": 0.4, "learning_rate": 0.0013665862237489251, "loss": 1.6445, "step": 4049 }, { "epoch": 0.4, "learning_rate": 0.0013662891334337015, "loss": 1.6416, "step": 4050 }, { "epoch": 0.4, "learning_rate": 0.0013659920057743902, "loss": 1.6045, "step": 4051 }, { "epoch": 0.4, "learning_rate": 0.001365694840801284, "loss": 1.4502, "step": 4052 }, { "epoch": 0.4, "learning_rate": 0.0013653976385446796, "loss": 1.6045, "step": 4053 }, { "epoch": 0.4, "learning_rate": 0.0013651003990348776, "loss": 1.5869, "step": 4054 }, { "epoch": 0.4, "learning_rate": 0.0013648031223021822, "loss": 1.5195, "step": 4055 }, { "epoch": 0.4, "learning_rate": 0.0013645058083769012, "loss": 1.5908, "step": 4056 }, { "epoch": 0.4, "learning_rate": 0.0013642084572893472, "loss": 1.5693, "step": 4057 }, { "epoch": 0.4, "learning_rate": 0.0013639110690698356, "loss": 1.6963, "step": 4058 }, { "epoch": 0.4, "learning_rate": 0.0013636136437486855, "loss": 1.6426, "step": 4059 }, { "epoch": 0.4, "learning_rate": 0.0013633161813562205, "loss": 1.5898, "step": 4060 }, { "epoch": 0.4, "learning_rate": 0.0013630186819227677, "loss": 1.4844, "step": 4061 }, { "epoch": 0.4, "learning_rate": 0.0013627211454786577, "loss": 1.6973, "step": 4062 }, { "epoch": 0.4, "learning_rate": 0.0013624235720542251, "loss": 1.6113, "step": 4063 }, { "epoch": 0.4, "learning_rate": 0.001362125961679808, "loss": 1.6064, "step": 4064 }, { "epoch": 0.4, "learning_rate": 0.0013618283143857492, "loss": 1.6279, "step": 4065 }, { "epoch": 0.4, "learning_rate": 0.0013615306302023943, "loss": 1.7393, "step": 4066 }, { "epoch": 0.4, "learning_rate": 0.0013612329091600927, "loss": 1.6699, "step": 4067 }, { "epoch": 0.4, "learning_rate": 0.0013609351512891977, "loss": 1.6562, "step": 4068 }, { "epoch": 0.4, "learning_rate": 0.0013606373566200669, "loss": 1.5889, "step": 4069 }, { "epoch": 0.4, "learning_rate": 0.0013603395251830612, "loss": 1.4824, "step": 4070 }, { "epoch": 0.4, "learning_rate": 0.0013600416570085448, "loss": 1.5938, "step": 4071 }, { "epoch": 0.4, "learning_rate": 0.0013597437521268864, "loss": 1.665, "step": 4072 }, { "epoch": 0.4, "learning_rate": 0.001359445810568458, "loss": 1.542, "step": 4073 }, { "epoch": 0.4, "learning_rate": 0.0013591478323636357, "loss": 1.625, "step": 4074 }, { "epoch": 0.4, "learning_rate": 0.001358849817542799, "loss": 1.6455, "step": 4075 }, { "epoch": 0.4, "learning_rate": 0.0013585517661363314, "loss": 1.748, "step": 4076 }, { "epoch": 0.4, "learning_rate": 0.0013582536781746198, "loss": 1.6699, "step": 4077 }, { "epoch": 0.4, "learning_rate": 0.001357955553688055, "loss": 1.6123, "step": 4078 }, { "epoch": 0.4, "learning_rate": 0.0013576573927070316, "loss": 1.6475, "step": 4079 }, { "epoch": 0.4, "learning_rate": 0.0013573591952619477, "loss": 1.5908, "step": 4080 }, { "epoch": 0.4, "learning_rate": 0.0013570609613832055, "loss": 1.6602, "step": 4081 }, { "epoch": 0.4, "learning_rate": 0.0013567626911012104, "loss": 1.5195, "step": 4082 }, { "epoch": 0.4, "learning_rate": 0.001356464384446372, "loss": 1.709, "step": 4083 }, { "epoch": 0.4, "learning_rate": 0.0013561660414491037, "loss": 1.6348, "step": 4084 }, { "epoch": 0.4, "learning_rate": 0.0013558676621398219, "loss": 1.5947, "step": 4085 }, { "epoch": 0.4, "learning_rate": 0.0013555692465489473, "loss": 1.5039, "step": 4086 }, { "epoch": 0.4, "learning_rate": 0.001355270794706904, "loss": 1.5781, "step": 4087 }, { "epoch": 0.4, "learning_rate": 0.00135497230664412, "loss": 1.6113, "step": 4088 }, { "epoch": 0.4, "learning_rate": 0.0013546737823910261, "loss": 1.543, "step": 4089 }, { "epoch": 0.4, "learning_rate": 0.001354375221978059, "loss": 1.5459, "step": 4090 }, { "epoch": 0.4, "learning_rate": 0.001354076625435657, "loss": 1.6123, "step": 4091 }, { "epoch": 0.4, "learning_rate": 0.0013537779927942626, "loss": 1.6436, "step": 4092 }, { "epoch": 0.4, "learning_rate": 0.0013534793240843224, "loss": 1.6543, "step": 4093 }, { "epoch": 0.4, "learning_rate": 0.0013531806193362863, "loss": 1.623, "step": 4094 }, { "epoch": 0.4, "learning_rate": 0.001352881878580608, "loss": 1.5908, "step": 4095 }, { "epoch": 0.4, "learning_rate": 0.0013525831018477447, "loss": 1.6104, "step": 4096 }, { "epoch": 0.4, "learning_rate": 0.001352284289168158, "loss": 1.6338, "step": 4097 }, { "epoch": 0.4, "learning_rate": 0.0013519854405723117, "loss": 1.5723, "step": 4098 }, { "epoch": 0.4, "learning_rate": 0.0013516865560906748, "loss": 1.4619, "step": 4099 }, { "epoch": 0.4, "learning_rate": 0.0013513876357537193, "loss": 1.585, "step": 4100 }, { "epoch": 0.4, "learning_rate": 0.0013510886795919208, "loss": 1.4717, "step": 4101 }, { "epoch": 0.4, "learning_rate": 0.0013507896876357587, "loss": 1.6064, "step": 4102 }, { "epoch": 0.4, "learning_rate": 0.0013504906599157154, "loss": 1.6582, "step": 4103 }, { "epoch": 0.4, "learning_rate": 0.0013501915964622782, "loss": 1.4912, "step": 4104 }, { "epoch": 0.4, "learning_rate": 0.0013498924973059372, "loss": 1.7021, "step": 4105 }, { "epoch": 0.4, "learning_rate": 0.0013495933624771861, "loss": 1.502, "step": 4106 }, { "epoch": 0.4, "learning_rate": 0.0013492941920065225, "loss": 1.6807, "step": 4107 }, { "epoch": 0.4, "learning_rate": 0.0013489949859244476, "loss": 1.6846, "step": 4108 }, { "epoch": 0.41, "learning_rate": 0.0013486957442614665, "loss": 1.4434, "step": 4109 }, { "epoch": 0.41, "learning_rate": 0.0013483964670480872, "loss": 1.6465, "step": 4110 }, { "epoch": 0.41, "learning_rate": 0.001348097154314822, "loss": 1.5576, "step": 4111 }, { "epoch": 0.41, "learning_rate": 0.0013477978060921865, "loss": 1.5674, "step": 4112 }, { "epoch": 0.41, "learning_rate": 0.0013474984224107, "loss": 1.7139, "step": 4113 }, { "epoch": 0.41, "learning_rate": 0.0013471990033008855, "loss": 1.4932, "step": 4114 }, { "epoch": 0.41, "learning_rate": 0.0013468995487932692, "loss": 1.5967, "step": 4115 }, { "epoch": 0.41, "learning_rate": 0.0013466000589183816, "loss": 1.584, "step": 4116 }, { "epoch": 0.41, "learning_rate": 0.0013463005337067565, "loss": 1.4648, "step": 4117 }, { "epoch": 0.41, "learning_rate": 0.0013460009731889307, "loss": 1.5654, "step": 4118 }, { "epoch": 0.41, "learning_rate": 0.0013457013773954456, "loss": 1.5273, "step": 4119 }, { "epoch": 0.41, "learning_rate": 0.0013454017463568454, "loss": 1.6787, "step": 4120 }, { "epoch": 0.41, "learning_rate": 0.001345102080103679, "loss": 1.5322, "step": 4121 }, { "epoch": 0.41, "learning_rate": 0.0013448023786664974, "loss": 1.5723, "step": 4122 }, { "epoch": 0.41, "learning_rate": 0.0013445026420758554, "loss": 1.6221, "step": 4123 }, { "epoch": 0.41, "learning_rate": 0.0013442028703623133, "loss": 1.5762, "step": 4124 }, { "epoch": 0.41, "learning_rate": 0.0013439030635564327, "loss": 1.5791, "step": 4125 }, { "epoch": 0.41, "learning_rate": 0.0013436032216887796, "loss": 1.6338, "step": 4126 }, { "epoch": 0.41, "learning_rate": 0.001343303344789924, "loss": 1.665, "step": 4127 }, { "epoch": 0.41, "learning_rate": 0.0013430034328904387, "loss": 1.6221, "step": 4128 }, { "epoch": 0.41, "learning_rate": 0.0013427034860209008, "loss": 1.6543, "step": 4129 }, { "epoch": 0.41, "learning_rate": 0.0013424035042118906, "loss": 1.6045, "step": 4130 }, { "epoch": 0.41, "learning_rate": 0.0013421034874939919, "loss": 1.5605, "step": 4131 }, { "epoch": 0.41, "learning_rate": 0.0013418034358977917, "loss": 1.5752, "step": 4132 }, { "epoch": 0.41, "learning_rate": 0.0013415033494538818, "loss": 1.6631, "step": 4133 }, { "epoch": 0.41, "learning_rate": 0.0013412032281928563, "loss": 1.624, "step": 4134 }, { "epoch": 0.41, "learning_rate": 0.0013409030721453137, "loss": 1.6191, "step": 4135 }, { "epoch": 0.41, "learning_rate": 0.0013406028813418552, "loss": 1.6064, "step": 4136 }, { "epoch": 0.41, "learning_rate": 0.001340302655813086, "loss": 1.626, "step": 4137 }, { "epoch": 0.41, "learning_rate": 0.0013400023955896152, "loss": 1.4961, "step": 4138 }, { "epoch": 0.41, "learning_rate": 0.0013397021007020549, "loss": 1.5176, "step": 4139 }, { "epoch": 0.41, "learning_rate": 0.0013394017711810208, "loss": 1.5254, "step": 4140 }, { "epoch": 0.41, "learning_rate": 0.0013391014070571325, "loss": 1.5098, "step": 4141 }, { "epoch": 0.41, "learning_rate": 0.0013388010083610121, "loss": 1.6211, "step": 4142 }, { "epoch": 0.41, "learning_rate": 0.0013385005751232872, "loss": 1.7344, "step": 4143 }, { "epoch": 0.41, "learning_rate": 0.001338200107374587, "loss": 1.6191, "step": 4144 }, { "epoch": 0.41, "learning_rate": 0.001337899605145545, "loss": 1.623, "step": 4145 }, { "epoch": 0.41, "learning_rate": 0.0013375990684667983, "loss": 1.6191, "step": 4146 }, { "epoch": 0.41, "learning_rate": 0.001337298497368987, "loss": 1.6406, "step": 4147 }, { "epoch": 0.41, "learning_rate": 0.0013369978918827557, "loss": 1.6484, "step": 4148 }, { "epoch": 0.41, "learning_rate": 0.001336697252038751, "loss": 1.6523, "step": 4149 }, { "epoch": 0.41, "learning_rate": 0.0013363965778676245, "loss": 1.5205, "step": 4150 }, { "epoch": 0.41, "learning_rate": 0.001336095869400031, "loss": 1.6533, "step": 4151 }, { "epoch": 0.41, "learning_rate": 0.0013357951266666278, "loss": 1.6592, "step": 4152 }, { "epoch": 0.41, "learning_rate": 0.0013354943496980765, "loss": 1.5898, "step": 4153 }, { "epoch": 0.41, "learning_rate": 0.0013351935385250422, "loss": 1.5059, "step": 4154 }, { "epoch": 0.41, "learning_rate": 0.0013348926931781935, "loss": 1.6113, "step": 4155 }, { "epoch": 0.41, "learning_rate": 0.001334591813688202, "loss": 1.6367, "step": 4156 }, { "epoch": 0.41, "learning_rate": 0.0013342909000857433, "loss": 1.5713, "step": 4157 }, { "epoch": 0.41, "learning_rate": 0.001333989952401496, "loss": 1.7734, "step": 4158 }, { "epoch": 0.41, "learning_rate": 0.001333688970666143, "loss": 1.7148, "step": 4159 }, { "epoch": 0.41, "learning_rate": 0.0013333879549103699, "loss": 1.6455, "step": 4160 }, { "epoch": 0.41, "learning_rate": 0.0013330869051648658, "loss": 1.5273, "step": 4161 }, { "epoch": 0.41, "learning_rate": 0.0013327858214603238, "loss": 1.6045, "step": 4162 }, { "epoch": 0.41, "learning_rate": 0.0013324847038274396, "loss": 1.5869, "step": 4163 }, { "epoch": 0.41, "learning_rate": 0.0013321835522969135, "loss": 1.6924, "step": 4164 }, { "epoch": 0.41, "learning_rate": 0.0013318823668994485, "loss": 1.54, "step": 4165 }, { "epoch": 0.41, "learning_rate": 0.0013315811476657506, "loss": 1.7227, "step": 4166 }, { "epoch": 0.41, "learning_rate": 0.0013312798946265307, "loss": 1.7285, "step": 4167 }, { "epoch": 0.41, "learning_rate": 0.001330978607812502, "loss": 1.5576, "step": 4168 }, { "epoch": 0.41, "learning_rate": 0.0013306772872543813, "loss": 1.6924, "step": 4169 }, { "epoch": 0.41, "learning_rate": 0.001330375932982889, "loss": 1.4355, "step": 4170 }, { "epoch": 0.41, "learning_rate": 0.0013300745450287492, "loss": 1.6924, "step": 4171 }, { "epoch": 0.41, "learning_rate": 0.0013297731234226888, "loss": 1.4922, "step": 4172 }, { "epoch": 0.41, "learning_rate": 0.001329471668195439, "loss": 1.5781, "step": 4173 }, { "epoch": 0.41, "learning_rate": 0.0013291701793777327, "loss": 1.7207, "step": 4174 }, { "epoch": 0.41, "learning_rate": 0.0013288686570003085, "loss": 1.666, "step": 4175 }, { "epoch": 0.41, "learning_rate": 0.0013285671010939074, "loss": 1.624, "step": 4176 }, { "epoch": 0.41, "learning_rate": 0.0013282655116892735, "loss": 1.5303, "step": 4177 }, { "epoch": 0.41, "learning_rate": 0.0013279638888171544, "loss": 1.5928, "step": 4178 }, { "epoch": 0.41, "learning_rate": 0.0013276622325083015, "loss": 1.7188, "step": 4179 }, { "epoch": 0.41, "learning_rate": 0.0013273605427934693, "loss": 1.5938, "step": 4180 }, { "epoch": 0.41, "learning_rate": 0.0013270588197034161, "loss": 1.459, "step": 4181 }, { "epoch": 0.41, "learning_rate": 0.0013267570632689032, "loss": 1.6182, "step": 4182 }, { "epoch": 0.41, "learning_rate": 0.0013264552735206948, "loss": 1.7002, "step": 4183 }, { "epoch": 0.41, "learning_rate": 0.00132615345048956, "loss": 1.5957, "step": 4184 }, { "epoch": 0.41, "learning_rate": 0.0013258515942062702, "loss": 1.5938, "step": 4185 }, { "epoch": 0.41, "learning_rate": 0.0013255497047016003, "loss": 1.6152, "step": 4186 }, { "epoch": 0.41, "learning_rate": 0.0013252477820063285, "loss": 1.6201, "step": 4187 }, { "epoch": 0.41, "learning_rate": 0.0013249458261512367, "loss": 1.6006, "step": 4188 }, { "epoch": 0.41, "learning_rate": 0.00132464383716711, "loss": 1.6768, "step": 4189 }, { "epoch": 0.41, "learning_rate": 0.0013243418150847373, "loss": 1.5566, "step": 4190 }, { "epoch": 0.41, "learning_rate": 0.0013240397599349099, "loss": 1.5762, "step": 4191 }, { "epoch": 0.41, "learning_rate": 0.0013237376717484233, "loss": 1.4951, "step": 4192 }, { "epoch": 0.41, "learning_rate": 0.0013234355505560764, "loss": 1.5, "step": 4193 }, { "epoch": 0.41, "learning_rate": 0.0013231333963886712, "loss": 1.5908, "step": 4194 }, { "epoch": 0.41, "learning_rate": 0.0013228312092770126, "loss": 1.5068, "step": 4195 }, { "epoch": 0.41, "learning_rate": 0.0013225289892519097, "loss": 1.5977, "step": 4196 }, { "epoch": 0.41, "learning_rate": 0.0013222267363441744, "loss": 1.6992, "step": 4197 }, { "epoch": 0.41, "learning_rate": 0.0013219244505846225, "loss": 1.6055, "step": 4198 }, { "epoch": 0.41, "learning_rate": 0.0013216221320040721, "loss": 1.5732, "step": 4199 }, { "epoch": 0.41, "learning_rate": 0.0013213197806333464, "loss": 1.7441, "step": 4200 }, { "epoch": 0.41, "learning_rate": 0.0013210173965032697, "loss": 1.6543, "step": 4201 }, { "epoch": 0.41, "learning_rate": 0.0013207149796446713, "loss": 1.5244, "step": 4202 }, { "epoch": 0.41, "learning_rate": 0.0013204125300883838, "loss": 1.5234, "step": 4203 }, { "epoch": 0.41, "learning_rate": 0.0013201100478652421, "loss": 1.6787, "step": 4204 }, { "epoch": 0.41, "learning_rate": 0.0013198075330060851, "loss": 1.5381, "step": 4205 }, { "epoch": 0.41, "learning_rate": 0.0013195049855417552, "loss": 1.4609, "step": 4206 }, { "epoch": 0.41, "learning_rate": 0.0013192024055030973, "loss": 1.6279, "step": 4207 }, { "epoch": 0.41, "learning_rate": 0.0013188997929209608, "loss": 1.4961, "step": 4208 }, { "epoch": 0.41, "learning_rate": 0.0013185971478261976, "loss": 1.7354, "step": 4209 }, { "epoch": 0.42, "learning_rate": 0.001318294470249663, "loss": 1.6426, "step": 4210 }, { "epoch": 0.42, "learning_rate": 0.001317991760222216, "loss": 1.7412, "step": 4211 }, { "epoch": 0.42, "learning_rate": 0.0013176890177747185, "loss": 1.6309, "step": 4212 }, { "epoch": 0.42, "learning_rate": 0.0013173862429380358, "loss": 1.5996, "step": 4213 }, { "epoch": 0.42, "learning_rate": 0.0013170834357430366, "loss": 1.6455, "step": 4214 }, { "epoch": 0.42, "learning_rate": 0.0013167805962205927, "loss": 1.7002, "step": 4215 }, { "epoch": 0.42, "learning_rate": 0.0013164777244015794, "loss": 1.5205, "step": 4216 }, { "epoch": 0.42, "learning_rate": 0.0013161748203168754, "loss": 1.5723, "step": 4217 }, { "epoch": 0.42, "learning_rate": 0.0013158718839973624, "loss": 1.5889, "step": 4218 }, { "epoch": 0.42, "learning_rate": 0.0013155689154739256, "loss": 1.5889, "step": 4219 }, { "epoch": 0.42, "learning_rate": 0.0013152659147774536, "loss": 1.5566, "step": 4220 }, { "epoch": 0.42, "learning_rate": 0.0013149628819388373, "loss": 1.7432, "step": 4221 }, { "epoch": 0.42, "learning_rate": 0.0013146598169889723, "loss": 1.5928, "step": 4222 }, { "epoch": 0.42, "learning_rate": 0.0013143567199587567, "loss": 1.7197, "step": 4223 }, { "epoch": 0.42, "learning_rate": 0.001314053590879092, "loss": 1.6416, "step": 4224 }, { "epoch": 0.42, "learning_rate": 0.001313750429780883, "loss": 1.5322, "step": 4225 }, { "epoch": 0.42, "learning_rate": 0.0013134472366950374, "loss": 1.541, "step": 4226 }, { "epoch": 0.42, "learning_rate": 0.0013131440116524672, "loss": 1.4336, "step": 4227 }, { "epoch": 0.42, "learning_rate": 0.001312840754684086, "loss": 1.665, "step": 4228 }, { "epoch": 0.42, "learning_rate": 0.0013125374658208124, "loss": 1.5752, "step": 4229 }, { "epoch": 0.42, "learning_rate": 0.0013122341450935674, "loss": 1.7227, "step": 4230 }, { "epoch": 0.42, "learning_rate": 0.0013119307925332747, "loss": 1.6533, "step": 4231 }, { "epoch": 0.42, "learning_rate": 0.0013116274081708622, "loss": 1.5596, "step": 4232 }, { "epoch": 0.42, "learning_rate": 0.0013113239920372606, "loss": 1.5527, "step": 4233 }, { "epoch": 0.42, "learning_rate": 0.0013110205441634042, "loss": 1.6064, "step": 4234 }, { "epoch": 0.42, "learning_rate": 0.00131071706458023, "loss": 1.6816, "step": 4235 }, { "epoch": 0.42, "learning_rate": 0.0013104135533186785, "loss": 1.5459, "step": 4236 }, { "epoch": 0.42, "learning_rate": 0.0013101100104096937, "loss": 1.5488, "step": 4237 }, { "epoch": 0.42, "learning_rate": 0.0013098064358842223, "loss": 1.6934, "step": 4238 }, { "epoch": 0.42, "learning_rate": 0.0013095028297732147, "loss": 1.5137, "step": 4239 }, { "epoch": 0.42, "learning_rate": 0.0013091991921076241, "loss": 1.5127, "step": 4240 }, { "epoch": 0.42, "learning_rate": 0.001308895522918407, "loss": 1.6475, "step": 4241 }, { "epoch": 0.42, "learning_rate": 0.0013085918222365232, "loss": 1.5859, "step": 4242 }, { "epoch": 0.42, "learning_rate": 0.001308288090092936, "loss": 1.71, "step": 4243 }, { "epoch": 0.42, "learning_rate": 0.001307984326518612, "loss": 1.6572, "step": 4244 }, { "epoch": 0.42, "learning_rate": 0.0013076805315445199, "loss": 1.5498, "step": 4245 }, { "epoch": 0.42, "learning_rate": 0.0013073767052016328, "loss": 1.5352, "step": 4246 }, { "epoch": 0.42, "learning_rate": 0.0013070728475209263, "loss": 1.5117, "step": 4247 }, { "epoch": 0.42, "learning_rate": 0.0013067689585333797, "loss": 1.5381, "step": 4248 }, { "epoch": 0.42, "learning_rate": 0.0013064650382699752, "loss": 1.6143, "step": 4249 }, { "epoch": 0.42, "learning_rate": 0.001306161086761698, "loss": 1.5303, "step": 4250 }, { "epoch": 0.42, "learning_rate": 0.0013058571040395366, "loss": 1.4922, "step": 4251 }, { "epoch": 0.42, "learning_rate": 0.0013055530901344835, "loss": 1.5361, "step": 4252 }, { "epoch": 0.42, "learning_rate": 0.001305249045077533, "loss": 1.666, "step": 4253 }, { "epoch": 0.42, "learning_rate": 0.0013049449688996835, "loss": 1.5967, "step": 4254 }, { "epoch": 0.42, "learning_rate": 0.0013046408616319365, "loss": 1.7188, "step": 4255 }, { "epoch": 0.42, "learning_rate": 0.001304336723305296, "loss": 1.6104, "step": 4256 }, { "epoch": 0.42, "learning_rate": 0.0013040325539507701, "loss": 1.4756, "step": 4257 }, { "epoch": 0.42, "learning_rate": 0.001303728353599369, "loss": 1.6143, "step": 4258 }, { "epoch": 0.42, "learning_rate": 0.0013034241222821077, "loss": 1.6348, "step": 4259 }, { "epoch": 0.42, "learning_rate": 0.0013031198600300024, "loss": 1.6104, "step": 4260 }, { "epoch": 0.42, "learning_rate": 0.0013028155668740743, "loss": 1.6816, "step": 4261 }, { "epoch": 0.42, "learning_rate": 0.0013025112428453457, "loss": 1.6484, "step": 4262 }, { "epoch": 0.42, "learning_rate": 0.0013022068879748442, "loss": 1.5986, "step": 4263 }, { "epoch": 0.42, "learning_rate": 0.001301902502293599, "loss": 1.6865, "step": 4264 }, { "epoch": 0.42, "learning_rate": 0.0013015980858326432, "loss": 1.5762, "step": 4265 }, { "epoch": 0.42, "learning_rate": 0.0013012936386230128, "loss": 1.4473, "step": 4266 }, { "epoch": 0.42, "learning_rate": 0.0013009891606957466, "loss": 1.6787, "step": 4267 }, { "epoch": 0.42, "learning_rate": 0.0013006846520818875, "loss": 1.6621, "step": 4268 }, { "epoch": 0.42, "learning_rate": 0.0013003801128124803, "loss": 1.5752, "step": 4269 }, { "epoch": 0.42, "learning_rate": 0.0013000755429185742, "loss": 1.5879, "step": 4270 }, { "epoch": 0.42, "learning_rate": 0.0012997709424312203, "loss": 1.6172, "step": 4271 }, { "epoch": 0.42, "learning_rate": 0.0012994663113814733, "loss": 1.4473, "step": 4272 }, { "epoch": 0.42, "learning_rate": 0.0012991616498003915, "loss": 1.5068, "step": 4273 }, { "epoch": 0.42, "learning_rate": 0.001298856957719036, "loss": 1.7129, "step": 4274 }, { "epoch": 0.42, "learning_rate": 0.0012985522351684703, "loss": 1.6416, "step": 4275 }, { "epoch": 0.42, "learning_rate": 0.001298247482179762, "loss": 1.5957, "step": 4276 }, { "epoch": 0.42, "learning_rate": 0.0012979426987839817, "loss": 1.6328, "step": 4277 }, { "epoch": 0.42, "learning_rate": 0.0012976378850122023, "loss": 1.6846, "step": 4278 }, { "epoch": 0.42, "learning_rate": 0.0012973330408955009, "loss": 1.6406, "step": 4279 }, { "epoch": 0.42, "learning_rate": 0.0012970281664649566, "loss": 1.6826, "step": 4280 }, { "epoch": 0.42, "learning_rate": 0.001296723261751652, "loss": 1.5518, "step": 4281 }, { "epoch": 0.42, "learning_rate": 0.0012964183267866734, "loss": 1.665, "step": 4282 }, { "epoch": 0.42, "learning_rate": 0.0012961133616011095, "loss": 1.7197, "step": 4283 }, { "epoch": 0.42, "learning_rate": 0.0012958083662260522, "loss": 1.5947, "step": 4284 }, { "epoch": 0.42, "learning_rate": 0.0012955033406925965, "loss": 1.75, "step": 4285 }, { "epoch": 0.42, "learning_rate": 0.0012951982850318406, "loss": 1.666, "step": 4286 }, { "epoch": 0.42, "learning_rate": 0.0012948931992748855, "loss": 1.6152, "step": 4287 }, { "epoch": 0.42, "learning_rate": 0.001294588083452836, "loss": 1.5635, "step": 4288 }, { "epoch": 0.42, "learning_rate": 0.0012942829375967987, "loss": 1.4492, "step": 4289 }, { "epoch": 0.42, "learning_rate": 0.0012939777617378843, "loss": 1.4355, "step": 4290 }, { "epoch": 0.42, "learning_rate": 0.001293672555907206, "loss": 1.5869, "step": 4291 }, { "epoch": 0.42, "learning_rate": 0.0012933673201358806, "loss": 1.6543, "step": 4292 }, { "epoch": 0.42, "learning_rate": 0.0012930620544550277, "loss": 1.6318, "step": 4293 }, { "epoch": 0.42, "learning_rate": 0.0012927567588957697, "loss": 1.5234, "step": 4294 }, { "epoch": 0.42, "learning_rate": 0.0012924514334892326, "loss": 1.5938, "step": 4295 }, { "epoch": 0.42, "learning_rate": 0.0012921460782665446, "loss": 1.4561, "step": 4296 }, { "epoch": 0.42, "learning_rate": 0.0012918406932588375, "loss": 1.5176, "step": 4297 }, { "epoch": 0.42, "learning_rate": 0.0012915352784972462, "loss": 1.5488, "step": 4298 }, { "epoch": 0.42, "learning_rate": 0.0012912298340129085, "loss": 1.5635, "step": 4299 }, { "epoch": 0.42, "learning_rate": 0.0012909243598369653, "loss": 1.6816, "step": 4300 }, { "epoch": 0.42, "learning_rate": 0.00129061885600056, "loss": 1.6455, "step": 4301 }, { "epoch": 0.42, "learning_rate": 0.00129031332253484, "loss": 1.6367, "step": 4302 }, { "epoch": 0.42, "learning_rate": 0.0012900077594709552, "loss": 1.7227, "step": 4303 }, { "epoch": 0.42, "learning_rate": 0.001289702166840058, "loss": 1.6631, "step": 4304 }, { "epoch": 0.42, "learning_rate": 0.001289396544673305, "loss": 1.6699, "step": 4305 }, { "epoch": 0.42, "learning_rate": 0.0012890908930018547, "loss": 1.6875, "step": 4306 }, { "epoch": 0.42, "learning_rate": 0.001288785211856869, "loss": 1.5801, "step": 4307 }, { "epoch": 0.42, "learning_rate": 0.0012884795012695132, "loss": 1.6025, "step": 4308 }, { "epoch": 0.42, "learning_rate": 0.001288173761270955, "loss": 1.6494, "step": 4309 }, { "epoch": 0.42, "learning_rate": 0.001287867991892365, "loss": 1.6865, "step": 4310 }, { "epoch": 0.42, "learning_rate": 0.0012875621931649182, "loss": 1.5645, "step": 4311 }, { "epoch": 0.43, "learning_rate": 0.0012872563651197906, "loss": 1.5137, "step": 4312 }, { "epoch": 0.43, "learning_rate": 0.0012869505077881625, "loss": 1.6172, "step": 4313 }, { "epoch": 0.43, "learning_rate": 0.0012866446212012165, "loss": 1.5908, "step": 4314 }, { "epoch": 0.43, "learning_rate": 0.001286338705390139, "loss": 1.5186, "step": 4315 }, { "epoch": 0.43, "learning_rate": 0.0012860327603861183, "loss": 1.5508, "step": 4316 }, { "epoch": 0.43, "learning_rate": 0.0012857267862203467, "loss": 1.5967, "step": 4317 }, { "epoch": 0.43, "learning_rate": 0.001285420782924019, "loss": 1.5039, "step": 4318 }, { "epoch": 0.43, "learning_rate": 0.001285114750528333, "loss": 1.7236, "step": 4319 }, { "epoch": 0.43, "learning_rate": 0.001284808689064489, "loss": 1.6641, "step": 4320 }, { "epoch": 0.43, "learning_rate": 0.0012845025985636913, "loss": 1.5439, "step": 4321 }, { "epoch": 0.43, "learning_rate": 0.0012841964790571464, "loss": 1.5811, "step": 4322 }, { "epoch": 0.43, "learning_rate": 0.001283890330576064, "loss": 1.7529, "step": 4323 }, { "epoch": 0.43, "learning_rate": 0.0012835841531516562, "loss": 1.4502, "step": 4324 }, { "epoch": 0.43, "learning_rate": 0.0012832779468151395, "loss": 1.5605, "step": 4325 }, { "epoch": 0.43, "learning_rate": 0.0012829717115977312, "loss": 1.5049, "step": 4326 }, { "epoch": 0.43, "learning_rate": 0.001282665447530654, "loss": 1.5947, "step": 4327 }, { "epoch": 0.43, "learning_rate": 0.0012823591546451314, "loss": 1.7588, "step": 4328 }, { "epoch": 0.43, "learning_rate": 0.001282052832972391, "loss": 1.5479, "step": 4329 }, { "epoch": 0.43, "learning_rate": 0.001281746482543663, "loss": 1.6855, "step": 4330 }, { "epoch": 0.43, "learning_rate": 0.001281440103390181, "loss": 1.6416, "step": 4331 }, { "epoch": 0.43, "learning_rate": 0.0012811336955431804, "loss": 1.5918, "step": 4332 }, { "epoch": 0.43, "learning_rate": 0.0012808272590339009, "loss": 1.4951, "step": 4333 }, { "epoch": 0.43, "learning_rate": 0.0012805207938935842, "loss": 1.665, "step": 4334 }, { "epoch": 0.43, "learning_rate": 0.0012802143001534746, "loss": 1.6191, "step": 4335 }, { "epoch": 0.43, "learning_rate": 0.001279907777844821, "loss": 1.4375, "step": 4336 }, { "epoch": 0.43, "learning_rate": 0.0012796012269988733, "loss": 1.5605, "step": 4337 }, { "epoch": 0.43, "learning_rate": 0.0012792946476468854, "loss": 1.6445, "step": 4338 }, { "epoch": 0.43, "learning_rate": 0.0012789880398201142, "loss": 1.6826, "step": 4339 }, { "epoch": 0.43, "learning_rate": 0.0012786814035498184, "loss": 1.5225, "step": 4340 }, { "epoch": 0.43, "learning_rate": 0.0012783747388672607, "loss": 1.585, "step": 4341 }, { "epoch": 0.43, "learning_rate": 0.0012780680458037061, "loss": 1.668, "step": 4342 }, { "epoch": 0.43, "learning_rate": 0.0012777613243904232, "loss": 1.7451, "step": 4343 }, { "epoch": 0.43, "learning_rate": 0.0012774545746586828, "loss": 1.5693, "step": 4344 }, { "epoch": 0.43, "learning_rate": 0.0012771477966397585, "loss": 1.4951, "step": 4345 }, { "epoch": 0.43, "learning_rate": 0.0012768409903649274, "loss": 1.4316, "step": 4346 }, { "epoch": 0.43, "learning_rate": 0.0012765341558654691, "loss": 1.6436, "step": 4347 }, { "epoch": 0.43, "learning_rate": 0.0012762272931726663, "loss": 1.6348, "step": 4348 }, { "epoch": 0.43, "learning_rate": 0.0012759204023178037, "loss": 1.5098, "step": 4349 }, { "epoch": 0.43, "learning_rate": 0.0012756134833321706, "loss": 1.5137, "step": 4350 }, { "epoch": 0.43, "learning_rate": 0.0012753065362470575, "loss": 1.6504, "step": 4351 }, { "epoch": 0.43, "learning_rate": 0.0012749995610937584, "loss": 1.5117, "step": 4352 }, { "epoch": 0.43, "learning_rate": 0.0012746925579035704, "loss": 1.5576, "step": 4353 }, { "epoch": 0.43, "learning_rate": 0.0012743855267077933, "loss": 1.5576, "step": 4354 }, { "epoch": 0.43, "learning_rate": 0.0012740784675377298, "loss": 1.623, "step": 4355 }, { "epoch": 0.43, "learning_rate": 0.001273771380424685, "loss": 1.4805, "step": 4356 }, { "epoch": 0.43, "learning_rate": 0.0012734642653999674, "loss": 1.5947, "step": 4357 }, { "epoch": 0.43, "learning_rate": 0.001273157122494888, "loss": 1.6953, "step": 4358 }, { "epoch": 0.43, "learning_rate": 0.0012728499517407608, "loss": 1.709, "step": 4359 }, { "epoch": 0.43, "learning_rate": 0.0012725427531689027, "loss": 1.5205, "step": 4360 }, { "epoch": 0.43, "learning_rate": 0.0012722355268106331, "loss": 1.541, "step": 4361 }, { "epoch": 0.43, "learning_rate": 0.001271928272697275, "loss": 1.502, "step": 4362 }, { "epoch": 0.43, "learning_rate": 0.0012716209908601535, "loss": 1.6826, "step": 4363 }, { "epoch": 0.43, "learning_rate": 0.0012713136813305966, "loss": 1.6514, "step": 4364 }, { "epoch": 0.43, "learning_rate": 0.0012710063441399356, "loss": 1.5283, "step": 4365 }, { "epoch": 0.43, "learning_rate": 0.0012706989793195035, "loss": 1.4502, "step": 4366 }, { "epoch": 0.43, "learning_rate": 0.0012703915869006382, "loss": 1.6074, "step": 4367 }, { "epoch": 0.43, "learning_rate": 0.001270084166914678, "loss": 1.5781, "step": 4368 }, { "epoch": 0.43, "learning_rate": 0.0012697767193929654, "loss": 1.4248, "step": 4369 }, { "epoch": 0.43, "learning_rate": 0.0012694692443668458, "loss": 1.6406, "step": 4370 }, { "epoch": 0.43, "learning_rate": 0.001269161741867667, "loss": 1.5293, "step": 4371 }, { "epoch": 0.43, "learning_rate": 0.001268854211926779, "loss": 1.5879, "step": 4372 }, { "epoch": 0.43, "learning_rate": 0.0012685466545755357, "loss": 1.5596, "step": 4373 }, { "epoch": 0.43, "learning_rate": 0.0012682390698452936, "loss": 1.7021, "step": 4374 }, { "epoch": 0.43, "learning_rate": 0.0012679314577674116, "loss": 1.6562, "step": 4375 }, { "epoch": 0.43, "learning_rate": 0.001267623818373251, "loss": 1.4629, "step": 4376 }, { "epoch": 0.43, "learning_rate": 0.0012673161516941764, "loss": 1.6172, "step": 4377 }, { "epoch": 0.43, "learning_rate": 0.0012670084577615563, "loss": 1.5889, "step": 4378 }, { "epoch": 0.43, "learning_rate": 0.0012667007366067597, "loss": 1.6572, "step": 4379 }, { "epoch": 0.43, "learning_rate": 0.0012663929882611603, "loss": 1.5283, "step": 4380 }, { "epoch": 0.43, "learning_rate": 0.0012660852127561332, "loss": 1.5723, "step": 4381 }, { "epoch": 0.43, "learning_rate": 0.0012657774101230573, "loss": 1.5703, "step": 4382 }, { "epoch": 0.43, "learning_rate": 0.0012654695803933135, "loss": 1.5762, "step": 4383 }, { "epoch": 0.43, "learning_rate": 0.0012651617235982862, "loss": 1.6006, "step": 4384 }, { "epoch": 0.43, "learning_rate": 0.0012648538397693617, "loss": 1.5059, "step": 4385 }, { "epoch": 0.43, "learning_rate": 0.0012645459289379298, "loss": 1.5303, "step": 4386 }, { "epoch": 0.43, "learning_rate": 0.0012642379911353828, "loss": 1.5527, "step": 4387 }, { "epoch": 0.43, "learning_rate": 0.0012639300263931157, "loss": 1.4629, "step": 4388 }, { "epoch": 0.43, "learning_rate": 0.0012636220347425264, "loss": 1.4795, "step": 4389 }, { "epoch": 0.43, "learning_rate": 0.0012633140162150152, "loss": 1.6152, "step": 4390 }, { "epoch": 0.43, "learning_rate": 0.0012630059708419851, "loss": 1.6406, "step": 4391 }, { "epoch": 0.43, "learning_rate": 0.001262697898654843, "loss": 1.6035, "step": 4392 }, { "epoch": 0.43, "learning_rate": 0.0012623897996849967, "loss": 1.7627, "step": 4393 }, { "epoch": 0.43, "learning_rate": 0.0012620816739638578, "loss": 1.4736, "step": 4394 }, { "epoch": 0.43, "learning_rate": 0.0012617735215228406, "loss": 1.498, "step": 4395 }, { "epoch": 0.43, "learning_rate": 0.0012614653423933628, "loss": 1.5215, "step": 4396 }, { "epoch": 0.43, "learning_rate": 0.0012611571366068428, "loss": 1.5996, "step": 4397 }, { "epoch": 0.43, "learning_rate": 0.0012608489041947034, "loss": 1.4883, "step": 4398 }, { "epoch": 0.43, "learning_rate": 0.00126054064518837, "loss": 1.5869, "step": 4399 }, { "epoch": 0.43, "learning_rate": 0.00126023235961927, "loss": 1.6201, "step": 4400 }, { "epoch": 0.43, "learning_rate": 0.0012599240475188339, "loss": 1.5332, "step": 4401 }, { "epoch": 0.43, "learning_rate": 0.0012596157089184948, "loss": 1.5137, "step": 4402 }, { "epoch": 0.43, "learning_rate": 0.0012593073438496886, "loss": 1.5791, "step": 4403 }, { "epoch": 0.43, "learning_rate": 0.0012589989523438545, "loss": 1.6543, "step": 4404 }, { "epoch": 0.43, "learning_rate": 0.001258690534432433, "loss": 1.4658, "step": 4405 }, { "epoch": 0.43, "learning_rate": 0.0012583820901468686, "loss": 1.5908, "step": 4406 }, { "epoch": 0.43, "learning_rate": 0.0012580736195186076, "loss": 1.5693, "step": 4407 }, { "epoch": 0.43, "learning_rate": 0.0012577651225790993, "loss": 1.5889, "step": 4408 }, { "epoch": 0.43, "learning_rate": 0.0012574565993597965, "loss": 1.5635, "step": 4409 }, { "epoch": 0.43, "learning_rate": 0.0012571480498921527, "loss": 1.4141, "step": 4410 }, { "epoch": 0.43, "learning_rate": 0.001256839474207626, "loss": 1.5889, "step": 4411 }, { "epoch": 0.43, "learning_rate": 0.0012565308723376765, "loss": 1.5938, "step": 4412 }, { "epoch": 0.44, "learning_rate": 0.001256222244313767, "loss": 1.4473, "step": 4413 }, { "epoch": 0.44, "learning_rate": 0.0012559135901673625, "loss": 1.6641, "step": 4414 }, { "epoch": 0.44, "learning_rate": 0.0012556049099299313, "loss": 1.584, "step": 4415 }, { "epoch": 0.44, "learning_rate": 0.001255296203632944, "loss": 1.6035, "step": 4416 }, { "epoch": 0.44, "learning_rate": 0.0012549874713078742, "loss": 1.4844, "step": 4417 }, { "epoch": 0.44, "learning_rate": 0.0012546787129861976, "loss": 1.5527, "step": 4418 }, { "epoch": 0.44, "learning_rate": 0.001254369928699393, "loss": 1.6348, "step": 4419 }, { "epoch": 0.44, "learning_rate": 0.0012540611184789416, "loss": 1.6055, "step": 4420 }, { "epoch": 0.44, "learning_rate": 0.0012537522823563281, "loss": 1.626, "step": 4421 }, { "epoch": 0.44, "learning_rate": 0.0012534434203630381, "loss": 1.5508, "step": 4422 }, { "epoch": 0.44, "learning_rate": 0.0012531345325305617, "loss": 1.5049, "step": 4423 }, { "epoch": 0.44, "learning_rate": 0.00125282561889039, "loss": 1.6797, "step": 4424 }, { "epoch": 0.44, "learning_rate": 0.0012525166794740185, "loss": 1.626, "step": 4425 }, { "epoch": 0.44, "learning_rate": 0.0012522077143129432, "loss": 1.6064, "step": 4426 }, { "epoch": 0.44, "learning_rate": 0.0012518987234386642, "loss": 1.583, "step": 4427 }, { "epoch": 0.44, "learning_rate": 0.0012515897068826845, "loss": 1.6191, "step": 4428 }, { "epoch": 0.44, "learning_rate": 0.0012512806646765085, "loss": 1.584, "step": 4429 }, { "epoch": 0.44, "learning_rate": 0.001250971596851644, "loss": 1.4883, "step": 4430 }, { "epoch": 0.44, "learning_rate": 0.0012506625034396017, "loss": 1.4766, "step": 4431 }, { "epoch": 0.44, "learning_rate": 0.0012503533844718936, "loss": 1.6895, "step": 4432 }, { "epoch": 0.44, "learning_rate": 0.0012500442399800357, "loss": 1.6025, "step": 4433 }, { "epoch": 0.44, "learning_rate": 0.001249735069995546, "loss": 1.5918, "step": 4434 }, { "epoch": 0.44, "learning_rate": 0.0012494258745499448, "loss": 1.5068, "step": 4435 }, { "epoch": 0.44, "learning_rate": 0.0012491166536747554, "loss": 1.5684, "step": 4436 }, { "epoch": 0.44, "learning_rate": 0.0012488074074015038, "loss": 1.6172, "step": 4437 }, { "epoch": 0.44, "learning_rate": 0.001248498135761719, "loss": 1.5693, "step": 4438 }, { "epoch": 0.44, "learning_rate": 0.001248188838786931, "loss": 1.6211, "step": 4439 }, { "epoch": 0.44, "learning_rate": 0.0012478795165086743, "loss": 1.4736, "step": 4440 }, { "epoch": 0.44, "learning_rate": 0.0012475701689584843, "loss": 1.6074, "step": 4441 }, { "epoch": 0.44, "learning_rate": 0.0012472607961679, "loss": 1.6582, "step": 4442 }, { "epoch": 0.44, "learning_rate": 0.001246951398168463, "loss": 1.7832, "step": 4443 }, { "epoch": 0.44, "learning_rate": 0.001246641974991717, "loss": 1.5977, "step": 4444 }, { "epoch": 0.44, "learning_rate": 0.0012463325266692082, "loss": 1.5889, "step": 4445 }, { "epoch": 0.44, "learning_rate": 0.0012460230532324863, "loss": 1.4082, "step": 4446 }, { "epoch": 0.44, "learning_rate": 0.0012457135547131023, "loss": 1.6592, "step": 4447 }, { "epoch": 0.44, "learning_rate": 0.0012454040311426105, "loss": 1.4883, "step": 4448 }, { "epoch": 0.44, "learning_rate": 0.0012450944825525678, "loss": 1.6055, "step": 4449 }, { "epoch": 0.44, "learning_rate": 0.001244784908974533, "loss": 1.6299, "step": 4450 }, { "epoch": 0.44, "learning_rate": 0.0012444753104400683, "loss": 1.6016, "step": 4451 }, { "epoch": 0.44, "learning_rate": 0.0012441656869807381, "loss": 1.5713, "step": 4452 }, { "epoch": 0.44, "learning_rate": 0.0012438560386281087, "loss": 1.5908, "step": 4453 }, { "epoch": 0.44, "learning_rate": 0.0012435463654137501, "loss": 1.5752, "step": 4454 }, { "epoch": 0.44, "learning_rate": 0.001243236667369234, "loss": 1.5127, "step": 4455 }, { "epoch": 0.44, "learning_rate": 0.0012429269445261354, "loss": 1.7168, "step": 4456 }, { "epoch": 0.44, "learning_rate": 0.0012426171969160303, "loss": 1.5615, "step": 4457 }, { "epoch": 0.44, "learning_rate": 0.001242307424570499, "loss": 1.6348, "step": 4458 }, { "epoch": 0.44, "learning_rate": 0.0012419976275211233, "loss": 1.5225, "step": 4459 }, { "epoch": 0.44, "learning_rate": 0.001241687805799488, "loss": 1.5996, "step": 4460 }, { "epoch": 0.44, "learning_rate": 0.0012413779594371795, "loss": 1.627, "step": 4461 }, { "epoch": 0.44, "learning_rate": 0.0012410680884657884, "loss": 1.6709, "step": 4462 }, { "epoch": 0.44, "learning_rate": 0.0012407581929169062, "loss": 1.6104, "step": 4463 }, { "epoch": 0.44, "learning_rate": 0.0012404482728221279, "loss": 1.6367, "step": 4464 }, { "epoch": 0.44, "learning_rate": 0.0012401383282130503, "loss": 1.7275, "step": 4465 }, { "epoch": 0.44, "learning_rate": 0.0012398283591212732, "loss": 1.6709, "step": 4466 }, { "epoch": 0.44, "learning_rate": 0.0012395183655783983, "loss": 1.6006, "step": 4467 }, { "epoch": 0.44, "learning_rate": 0.001239208347616031, "loss": 1.5098, "step": 4468 }, { "epoch": 0.44, "learning_rate": 0.0012388983052657777, "loss": 1.4863, "step": 4469 }, { "epoch": 0.44, "learning_rate": 0.001238588238559248, "loss": 1.4727, "step": 4470 }, { "epoch": 0.44, "learning_rate": 0.0012382781475280544, "loss": 1.7549, "step": 4471 }, { "epoch": 0.44, "learning_rate": 0.0012379680322038115, "loss": 1.4424, "step": 4472 }, { "epoch": 0.44, "learning_rate": 0.0012376578926181362, "loss": 1.6855, "step": 4473 }, { "epoch": 0.44, "learning_rate": 0.0012373477288026475, "loss": 1.5234, "step": 4474 }, { "epoch": 0.44, "learning_rate": 0.001237037540788968, "loss": 1.5771, "step": 4475 }, { "epoch": 0.44, "learning_rate": 0.001236727328608722, "loss": 1.6816, "step": 4476 }, { "epoch": 0.44, "learning_rate": 0.0012364170922935362, "loss": 1.624, "step": 4477 }, { "epoch": 0.44, "learning_rate": 0.00123610683187504, "loss": 1.6152, "step": 4478 }, { "epoch": 0.44, "learning_rate": 0.001235796547384865, "loss": 1.6289, "step": 4479 }, { "epoch": 0.44, "learning_rate": 0.0012354862388546466, "loss": 1.752, "step": 4480 }, { "epoch": 0.44, "learning_rate": 0.00123517590631602, "loss": 1.6094, "step": 4481 }, { "epoch": 0.44, "learning_rate": 0.0012348655498006255, "loss": 1.582, "step": 4482 }, { "epoch": 0.44, "learning_rate": 0.0012345551693401042, "loss": 1.6006, "step": 4483 }, { "epoch": 0.44, "learning_rate": 0.0012342447649661002, "loss": 1.5801, "step": 4484 }, { "epoch": 0.44, "learning_rate": 0.00123393433671026, "loss": 1.4888, "step": 4485 }, { "epoch": 0.44, "learning_rate": 0.0012336238846042324, "loss": 1.5635, "step": 4486 }, { "epoch": 0.44, "learning_rate": 0.001233313408679669, "loss": 1.5752, "step": 4487 }, { "epoch": 0.44, "learning_rate": 0.001233002908968224, "loss": 1.5576, "step": 4488 }, { "epoch": 0.44, "learning_rate": 0.0012326923855015528, "loss": 1.6338, "step": 4489 }, { "epoch": 0.44, "learning_rate": 0.0012323818383113143, "loss": 1.543, "step": 4490 }, { "epoch": 0.44, "learning_rate": 0.00123207126742917, "loss": 1.5566, "step": 4491 }, { "epoch": 0.44, "learning_rate": 0.0012317606728867827, "loss": 1.6274, "step": 4492 }, { "epoch": 0.44, "learning_rate": 0.0012314500547158187, "loss": 1.5527, "step": 4493 }, { "epoch": 0.44, "learning_rate": 0.0012311394129479463, "loss": 1.6611, "step": 4494 }, { "epoch": 0.44, "learning_rate": 0.001230828747614836, "loss": 1.5479, "step": 4495 }, { "epoch": 0.44, "learning_rate": 0.001230518058748161, "loss": 1.5928, "step": 4496 }, { "epoch": 0.44, "learning_rate": 0.0012302073463795968, "loss": 1.6416, "step": 4497 }, { "epoch": 0.44, "learning_rate": 0.0012298966105408215, "loss": 1.4971, "step": 4498 }, { "epoch": 0.44, "learning_rate": 0.0012295858512635152, "loss": 1.5791, "step": 4499 }, { "epoch": 0.44, "learning_rate": 0.0012292750685793604, "loss": 1.5674, "step": 4500 }, { "epoch": 0.44, "learning_rate": 0.0012289642625200425, "loss": 1.5439, "step": 4501 }, { "epoch": 0.44, "learning_rate": 0.0012286534331172488, "loss": 1.5195, "step": 4502 }, { "epoch": 0.44, "learning_rate": 0.0012283425804026692, "loss": 1.6357, "step": 4503 }, { "epoch": 0.44, "learning_rate": 0.0012280317044079953, "loss": 1.5938, "step": 4504 }, { "epoch": 0.44, "learning_rate": 0.0012277208051649227, "loss": 1.499, "step": 4505 }, { "epoch": 0.44, "learning_rate": 0.001227409882705148, "loss": 1.5254, "step": 4506 }, { "epoch": 0.44, "learning_rate": 0.00122709893706037, "loss": 1.54, "step": 4507 }, { "epoch": 0.44, "learning_rate": 0.0012267879682622913, "loss": 1.6855, "step": 4508 }, { "epoch": 0.44, "learning_rate": 0.001226476976342615, "loss": 1.6221, "step": 4509 }, { "epoch": 0.44, "learning_rate": 0.001226165961333048, "loss": 1.6416, "step": 4510 }, { "epoch": 0.44, "learning_rate": 0.0012258549232652988, "loss": 1.5146, "step": 4511 }, { "epoch": 0.44, "learning_rate": 0.0012255438621710788, "loss": 1.5947, "step": 4512 }, { "epoch": 0.44, "learning_rate": 0.0012252327780821015, "loss": 1.6084, "step": 4513 }, { "epoch": 0.44, "learning_rate": 0.0012249216710300823, "loss": 1.5811, "step": 4514 }, { "epoch": 0.45, "learning_rate": 0.0012246105410467396, "loss": 1.624, "step": 4515 }, { "epoch": 0.45, "learning_rate": 0.001224299388163794, "loss": 1.709, "step": 4516 }, { "epoch": 0.45, "learning_rate": 0.001223988212412968, "loss": 1.6436, "step": 4517 }, { "epoch": 0.45, "learning_rate": 0.0012236770138259869, "loss": 1.4639, "step": 4518 }, { "epoch": 0.45, "learning_rate": 0.0012233657924345782, "loss": 1.5059, "step": 4519 }, { "epoch": 0.45, "learning_rate": 0.0012230545482704715, "loss": 1.6416, "step": 4520 }, { "epoch": 0.45, "learning_rate": 0.001222743281365399, "loss": 1.6074, "step": 4521 }, { "epoch": 0.45, "learning_rate": 0.0012224319917510953, "loss": 1.5742, "step": 4522 }, { "epoch": 0.45, "learning_rate": 0.0012221206794592971, "loss": 1.6426, "step": 4523 }, { "epoch": 0.45, "learning_rate": 0.0012218093445217435, "loss": 1.5752, "step": 4524 }, { "epoch": 0.45, "learning_rate": 0.0012214979869701757, "loss": 1.4561, "step": 4525 }, { "epoch": 0.45, "learning_rate": 0.0012211866068363374, "loss": 1.6699, "step": 4526 }, { "epoch": 0.45, "learning_rate": 0.0012208752041519748, "loss": 1.6094, "step": 4527 }, { "epoch": 0.45, "learning_rate": 0.001220563778948836, "loss": 1.5303, "step": 4528 }, { "epoch": 0.45, "learning_rate": 0.0012202523312586711, "loss": 1.6338, "step": 4529 }, { "epoch": 0.45, "learning_rate": 0.001219940861113234, "loss": 1.5811, "step": 4530 }, { "epoch": 0.45, "learning_rate": 0.001219629368544279, "loss": 1.6035, "step": 4531 }, { "epoch": 0.45, "learning_rate": 0.001219317853583564, "loss": 1.5996, "step": 4532 }, { "epoch": 0.45, "learning_rate": 0.0012190063162628487, "loss": 1.5293, "step": 4533 }, { "epoch": 0.45, "learning_rate": 0.0012186947566138948, "loss": 1.5264, "step": 4534 }, { "epoch": 0.45, "learning_rate": 0.0012183831746684667, "loss": 1.5508, "step": 4535 }, { "epoch": 0.45, "learning_rate": 0.0012180715704583314, "loss": 1.5869, "step": 4536 }, { "epoch": 0.45, "learning_rate": 0.001217759944015257, "loss": 1.4893, "step": 4537 }, { "epoch": 0.45, "learning_rate": 0.0012174482953710149, "loss": 1.6035, "step": 4538 }, { "epoch": 0.45, "learning_rate": 0.0012171366245573786, "loss": 1.5967, "step": 4539 }, { "epoch": 0.45, "learning_rate": 0.0012168249316061237, "loss": 1.6982, "step": 4540 }, { "epoch": 0.45, "learning_rate": 0.0012165132165490279, "loss": 1.5303, "step": 4541 }, { "epoch": 0.45, "learning_rate": 0.0012162014794178714, "loss": 1.5557, "step": 4542 }, { "epoch": 0.45, "learning_rate": 0.0012158897202444363, "loss": 1.5508, "step": 4543 }, { "epoch": 0.45, "learning_rate": 0.0012155779390605078, "loss": 1.5615, "step": 4544 }, { "epoch": 0.45, "learning_rate": 0.0012152661358978722, "loss": 1.501, "step": 4545 }, { "epoch": 0.45, "learning_rate": 0.001214954310788319, "loss": 1.4717, "step": 4546 }, { "epoch": 0.45, "learning_rate": 0.0012146424637636396, "loss": 1.6406, "step": 4547 }, { "epoch": 0.45, "learning_rate": 0.0012143305948556272, "loss": 1.5566, "step": 4548 }, { "epoch": 0.45, "learning_rate": 0.0012140187040960778, "loss": 1.7021, "step": 4549 }, { "epoch": 0.45, "learning_rate": 0.0012137067915167896, "loss": 1.5771, "step": 4550 }, { "epoch": 0.45, "learning_rate": 0.0012133948571495625, "loss": 1.625, "step": 4551 }, { "epoch": 0.45, "learning_rate": 0.0012130829010261995, "loss": 1.6182, "step": 4552 }, { "epoch": 0.45, "learning_rate": 0.0012127709231785047, "loss": 1.6211, "step": 4553 }, { "epoch": 0.45, "learning_rate": 0.001212458923638285, "loss": 1.5957, "step": 4554 }, { "epoch": 0.45, "learning_rate": 0.0012121469024373504, "loss": 1.542, "step": 4555 }, { "epoch": 0.45, "learning_rate": 0.0012118348596075114, "loss": 1.583, "step": 4556 }, { "epoch": 0.45, "learning_rate": 0.0012115227951805818, "loss": 1.582, "step": 4557 }, { "epoch": 0.45, "learning_rate": 0.0012112107091883774, "loss": 1.75, "step": 4558 }, { "epoch": 0.45, "learning_rate": 0.0012108986016627164, "loss": 1.501, "step": 4559 }, { "epoch": 0.45, "learning_rate": 0.0012105864726354183, "loss": 1.5684, "step": 4560 }, { "epoch": 0.45, "learning_rate": 0.001210274322138306, "loss": 1.5918, "step": 4561 }, { "epoch": 0.45, "learning_rate": 0.0012099621502032036, "loss": 1.5205, "step": 4562 }, { "epoch": 0.45, "learning_rate": 0.001209649956861938, "loss": 1.6143, "step": 4563 }, { "epoch": 0.45, "learning_rate": 0.0012093377421463383, "loss": 1.665, "step": 4564 }, { "epoch": 0.45, "learning_rate": 0.0012090255060882352, "loss": 1.5615, "step": 4565 }, { "epoch": 0.45, "learning_rate": 0.001208713248719462, "loss": 1.5693, "step": 4566 }, { "epoch": 0.45, "learning_rate": 0.0012084009700718543, "loss": 1.6201, "step": 4567 }, { "epoch": 0.45, "learning_rate": 0.0012080886701772498, "loss": 1.7393, "step": 4568 }, { "epoch": 0.45, "learning_rate": 0.0012077763490674878, "loss": 1.584, "step": 4569 }, { "epoch": 0.45, "learning_rate": 0.0012074640067744106, "loss": 1.7168, "step": 4570 }, { "epoch": 0.45, "learning_rate": 0.0012071516433298617, "loss": 1.6133, "step": 4571 }, { "epoch": 0.45, "learning_rate": 0.001206839258765688, "loss": 1.5215, "step": 4572 }, { "epoch": 0.45, "learning_rate": 0.0012065268531137377, "loss": 1.6934, "step": 4573 }, { "epoch": 0.45, "learning_rate": 0.0012062144264058615, "loss": 1.6016, "step": 4574 }, { "epoch": 0.45, "learning_rate": 0.0012059019786739113, "loss": 1.5576, "step": 4575 }, { "epoch": 0.45, "learning_rate": 0.001205589509949743, "loss": 1.6816, "step": 4576 }, { "epoch": 0.45, "learning_rate": 0.0012052770202652127, "loss": 1.5049, "step": 4577 }, { "epoch": 0.45, "learning_rate": 0.0012049645096521803, "loss": 1.6045, "step": 4578 }, { "epoch": 0.45, "learning_rate": 0.0012046519781425062, "loss": 1.6836, "step": 4579 }, { "epoch": 0.45, "learning_rate": 0.001204339425768054, "loss": 1.6943, "step": 4580 }, { "epoch": 0.45, "learning_rate": 0.0012040268525606895, "loss": 1.6738, "step": 4581 }, { "epoch": 0.45, "learning_rate": 0.0012037142585522803, "loss": 1.7041, "step": 4582 }, { "epoch": 0.45, "learning_rate": 0.0012034016437746957, "loss": 1.6055, "step": 4583 }, { "epoch": 0.45, "learning_rate": 0.001203089008259808, "loss": 1.5654, "step": 4584 }, { "epoch": 0.45, "learning_rate": 0.001202776352039491, "loss": 1.7695, "step": 4585 }, { "epoch": 0.45, "learning_rate": 0.0012024636751456208, "loss": 1.6348, "step": 4586 }, { "epoch": 0.45, "learning_rate": 0.0012021509776100757, "loss": 1.6475, "step": 4587 }, { "epoch": 0.45, "learning_rate": 0.0012018382594647357, "loss": 1.5889, "step": 4588 }, { "epoch": 0.45, "learning_rate": 0.0012015255207414833, "loss": 1.4834, "step": 4589 }, { "epoch": 0.45, "learning_rate": 0.0012012127614722031, "loss": 1.6572, "step": 4590 }, { "epoch": 0.45, "learning_rate": 0.0012008999816887815, "loss": 1.6787, "step": 4591 }, { "epoch": 0.45, "learning_rate": 0.0012005871814231078, "loss": 1.5557, "step": 4592 }, { "epoch": 0.45, "learning_rate": 0.001200274360707072, "loss": 1.6797, "step": 4593 }, { "epoch": 0.45, "learning_rate": 0.001199961519572567, "loss": 1.6748, "step": 4594 }, { "epoch": 0.45, "learning_rate": 0.001199648658051488, "loss": 1.499, "step": 4595 }, { "epoch": 0.45, "learning_rate": 0.001199335776175732, "loss": 1.5635, "step": 4596 }, { "epoch": 0.45, "learning_rate": 0.001199022873977198, "loss": 1.6221, "step": 4597 }, { "epoch": 0.45, "learning_rate": 0.0011987099514877873, "loss": 1.5342, "step": 4598 }, { "epoch": 0.45, "learning_rate": 0.0011983970087394032, "loss": 1.5811, "step": 4599 }, { "epoch": 0.45, "learning_rate": 0.0011980840457639507, "loss": 1.6689, "step": 4600 }, { "epoch": 0.45, "learning_rate": 0.001197771062593337, "loss": 1.46, "step": 4601 }, { "epoch": 0.45, "learning_rate": 0.001197458059259472, "loss": 1.6113, "step": 4602 }, { "epoch": 0.45, "learning_rate": 0.001197145035794267, "loss": 1.5908, "step": 4603 }, { "epoch": 0.45, "learning_rate": 0.0011968319922296357, "loss": 1.5752, "step": 4604 }, { "epoch": 0.45, "learning_rate": 0.0011965189285974929, "loss": 1.6475, "step": 4605 }, { "epoch": 0.45, "learning_rate": 0.0011962058449297571, "loss": 1.6338, "step": 4606 }, { "epoch": 0.45, "learning_rate": 0.0011958927412583473, "loss": 1.6221, "step": 4607 }, { "epoch": 0.45, "learning_rate": 0.0011955796176151858, "loss": 1.5654, "step": 4608 }, { "epoch": 0.45, "learning_rate": 0.001195266474032196, "loss": 1.6104, "step": 4609 }, { "epoch": 0.45, "learning_rate": 0.001194953310541304, "loss": 1.5674, "step": 4610 }, { "epoch": 0.45, "learning_rate": 0.0011946401271744368, "loss": 1.6494, "step": 4611 }, { "epoch": 0.45, "learning_rate": 0.001194326923963525, "loss": 1.5967, "step": 4612 }, { "epoch": 0.45, "learning_rate": 0.0011940137009405004, "loss": 1.666, "step": 4613 }, { "epoch": 0.45, "learning_rate": 0.0011937004581372961, "loss": 1.6152, "step": 4614 }, { "epoch": 0.45, "learning_rate": 0.001193387195585849, "loss": 1.6533, "step": 4615 }, { "epoch": 0.46, "learning_rate": 0.0011930739133180963, "loss": 1.582, "step": 4616 }, { "epoch": 0.46, "learning_rate": 0.0011927606113659782, "loss": 1.4668, "step": 4617 }, { "epoch": 0.46, "learning_rate": 0.0011924472897614368, "loss": 1.583, "step": 4618 }, { "epoch": 0.46, "learning_rate": 0.0011921339485364154, "loss": 1.4111, "step": 4619 }, { "epoch": 0.46, "learning_rate": 0.0011918205877228603, "loss": 1.5303, "step": 4620 }, { "epoch": 0.46, "learning_rate": 0.0011915072073527195, "loss": 1.6943, "step": 4621 }, { "epoch": 0.46, "learning_rate": 0.0011911938074579426, "loss": 1.6279, "step": 4622 }, { "epoch": 0.46, "learning_rate": 0.0011908803880704818, "loss": 1.6104, "step": 4623 }, { "epoch": 0.46, "learning_rate": 0.0011905669492222912, "loss": 1.7285, "step": 4624 }, { "epoch": 0.46, "learning_rate": 0.0011902534909453261, "loss": 1.4756, "step": 4625 }, { "epoch": 0.46, "learning_rate": 0.0011899400132715447, "loss": 1.5029, "step": 4626 }, { "epoch": 0.46, "learning_rate": 0.0011896265162329067, "loss": 1.5713, "step": 4627 }, { "epoch": 0.46, "learning_rate": 0.001189312999861374, "loss": 1.5215, "step": 4628 }, { "epoch": 0.46, "learning_rate": 0.0011889994641889102, "loss": 1.6426, "step": 4629 }, { "epoch": 0.46, "learning_rate": 0.0011886859092474808, "loss": 1.6035, "step": 4630 }, { "epoch": 0.46, "learning_rate": 0.0011883723350690544, "loss": 1.582, "step": 4631 }, { "epoch": 0.46, "learning_rate": 0.0011880587416855997, "loss": 1.4678, "step": 4632 }, { "epoch": 0.46, "learning_rate": 0.001187745129129089, "loss": 1.5869, "step": 4633 }, { "epoch": 0.46, "learning_rate": 0.0011874314974314956, "loss": 1.4365, "step": 4634 }, { "epoch": 0.46, "learning_rate": 0.0011871178466247951, "loss": 1.4443, "step": 4635 }, { "epoch": 0.46, "learning_rate": 0.001186804176740965, "loss": 1.5576, "step": 4636 }, { "epoch": 0.46, "learning_rate": 0.0011864904878119847, "loss": 1.5889, "step": 4637 }, { "epoch": 0.46, "learning_rate": 0.0011861767798698353, "loss": 1.5635, "step": 4638 }, { "epoch": 0.46, "learning_rate": 0.0011858630529465002, "loss": 1.6328, "step": 4639 }, { "epoch": 0.46, "learning_rate": 0.001185549307073965, "loss": 1.6025, "step": 4640 }, { "epoch": 0.46, "learning_rate": 0.0011852355422842166, "loss": 1.6533, "step": 4641 }, { "epoch": 0.46, "learning_rate": 0.0011849217586092442, "loss": 1.6572, "step": 4642 }, { "epoch": 0.46, "learning_rate": 0.0011846079560810388, "loss": 1.5449, "step": 4643 }, { "epoch": 0.46, "learning_rate": 0.0011842941347315935, "loss": 1.5107, "step": 4644 }, { "epoch": 0.46, "learning_rate": 0.0011839802945929026, "loss": 1.6074, "step": 4645 }, { "epoch": 0.46, "learning_rate": 0.0011836664356969636, "loss": 1.5732, "step": 4646 }, { "epoch": 0.46, "learning_rate": 0.0011833525580757748, "loss": 1.668, "step": 4647 }, { "epoch": 0.46, "learning_rate": 0.0011830386617613365, "loss": 1.4854, "step": 4648 }, { "epoch": 0.46, "learning_rate": 0.001182724746785652, "loss": 1.5146, "step": 4649 }, { "epoch": 0.46, "learning_rate": 0.0011824108131807253, "loss": 1.6025, "step": 4650 }, { "epoch": 0.46, "learning_rate": 0.0011820968609785622, "loss": 1.7129, "step": 4651 }, { "epoch": 0.46, "learning_rate": 0.001181782890211172, "loss": 1.584, "step": 4652 }, { "epoch": 0.46, "learning_rate": 0.001181468900910564, "loss": 1.6582, "step": 4653 }, { "epoch": 0.46, "learning_rate": 0.0011811548931087504, "loss": 1.6006, "step": 4654 }, { "epoch": 0.46, "learning_rate": 0.0011808408668377447, "loss": 1.4854, "step": 4655 }, { "epoch": 0.46, "learning_rate": 0.0011805268221295633, "loss": 1.5479, "step": 4656 }, { "epoch": 0.46, "learning_rate": 0.0011802127590162237, "loss": 1.6836, "step": 4657 }, { "epoch": 0.46, "learning_rate": 0.0011798986775297454, "loss": 1.6543, "step": 4658 }, { "epoch": 0.46, "learning_rate": 0.0011795845777021494, "loss": 1.6016, "step": 4659 }, { "epoch": 0.46, "learning_rate": 0.0011792704595654593, "loss": 1.6172, "step": 4660 }, { "epoch": 0.46, "learning_rate": 0.0011789563231517002, "loss": 1.6064, "step": 4661 }, { "epoch": 0.46, "learning_rate": 0.0011786421684928992, "loss": 1.6182, "step": 4662 }, { "epoch": 0.46, "learning_rate": 0.0011783279956210846, "loss": 1.6191, "step": 4663 }, { "epoch": 0.46, "learning_rate": 0.0011780138045682875, "loss": 1.6094, "step": 4664 }, { "epoch": 0.46, "learning_rate": 0.0011776995953665407, "loss": 1.6572, "step": 4665 }, { "epoch": 0.46, "learning_rate": 0.0011773853680478786, "loss": 1.5596, "step": 4666 }, { "epoch": 0.46, "learning_rate": 0.001177071122644337, "loss": 1.6436, "step": 4667 }, { "epoch": 0.46, "learning_rate": 0.0011767568591879542, "loss": 1.5049, "step": 4668 }, { "epoch": 0.46, "learning_rate": 0.00117644257771077, "loss": 1.5967, "step": 4669 }, { "epoch": 0.46, "learning_rate": 0.0011761282782448265, "loss": 1.6562, "step": 4670 }, { "epoch": 0.46, "learning_rate": 0.0011758139608221674, "loss": 1.5547, "step": 4671 }, { "epoch": 0.46, "learning_rate": 0.0011754996254748377, "loss": 1.6748, "step": 4672 }, { "epoch": 0.46, "learning_rate": 0.0011751852722348846, "loss": 1.5635, "step": 4673 }, { "epoch": 0.46, "learning_rate": 0.0011748709011343576, "loss": 1.5615, "step": 4674 }, { "epoch": 0.46, "learning_rate": 0.001174556512205308, "loss": 1.5146, "step": 4675 }, { "epoch": 0.46, "learning_rate": 0.0011742421054797875, "loss": 1.5312, "step": 4676 }, { "epoch": 0.46, "learning_rate": 0.0011739276809898513, "loss": 1.6309, "step": 4677 }, { "epoch": 0.46, "learning_rate": 0.0011736132387675559, "loss": 1.5186, "step": 4678 }, { "epoch": 0.46, "learning_rate": 0.0011732987788449587, "loss": 1.6396, "step": 4679 }, { "epoch": 0.46, "learning_rate": 0.0011729843012541201, "loss": 1.5342, "step": 4680 }, { "epoch": 0.46, "learning_rate": 0.0011726698060271021, "loss": 1.6475, "step": 4681 }, { "epoch": 0.46, "learning_rate": 0.0011723552931959683, "loss": 1.5635, "step": 4682 }, { "epoch": 0.46, "learning_rate": 0.0011720407627927836, "loss": 1.4766, "step": 4683 }, { "epoch": 0.46, "learning_rate": 0.0011717262148496156, "loss": 1.6104, "step": 4684 }, { "epoch": 0.46, "learning_rate": 0.0011714116493985332, "loss": 1.4155, "step": 4685 }, { "epoch": 0.46, "learning_rate": 0.0011710970664716067, "loss": 1.6611, "step": 4686 }, { "epoch": 0.46, "learning_rate": 0.001170782466100909, "loss": 1.5635, "step": 4687 }, { "epoch": 0.46, "learning_rate": 0.0011704678483185143, "loss": 1.5068, "step": 4688 }, { "epoch": 0.46, "learning_rate": 0.0011701532131564985, "loss": 1.5801, "step": 4689 }, { "epoch": 0.46, "learning_rate": 0.0011698385606469398, "loss": 1.6475, "step": 4690 }, { "epoch": 0.46, "learning_rate": 0.0011695238908219175, "loss": 1.4268, "step": 4691 }, { "epoch": 0.46, "learning_rate": 0.0011692092037135131, "loss": 1.5332, "step": 4692 }, { "epoch": 0.46, "learning_rate": 0.0011688944993538098, "loss": 1.4902, "step": 4693 }, { "epoch": 0.46, "learning_rate": 0.0011685797777748923, "loss": 1.6914, "step": 4694 }, { "epoch": 0.46, "learning_rate": 0.0011682650390088473, "loss": 1.7461, "step": 4695 }, { "epoch": 0.46, "learning_rate": 0.0011679502830877634, "loss": 1.6934, "step": 4696 }, { "epoch": 0.46, "learning_rate": 0.0011676355100437305, "loss": 1.5615, "step": 4697 }, { "epoch": 0.46, "learning_rate": 0.0011673207199088406, "loss": 1.6279, "step": 4698 }, { "epoch": 0.46, "learning_rate": 0.001167005912715187, "loss": 1.6982, "step": 4699 }, { "epoch": 0.46, "learning_rate": 0.0011666910884948658, "loss": 1.5264, "step": 4700 }, { "epoch": 0.46, "learning_rate": 0.0011663762472799736, "loss": 1.5703, "step": 4701 }, { "epoch": 0.46, "learning_rate": 0.0011660613891026094, "loss": 1.5957, "step": 4702 }, { "epoch": 0.46, "learning_rate": 0.001165746513994874, "loss": 1.4297, "step": 4703 }, { "epoch": 0.46, "learning_rate": 0.0011654316219888692, "loss": 1.6211, "step": 4704 }, { "epoch": 0.46, "learning_rate": 0.001165116713116699, "loss": 1.4707, "step": 4705 }, { "epoch": 0.46, "learning_rate": 0.0011648017874104698, "loss": 1.6592, "step": 4706 }, { "epoch": 0.46, "learning_rate": 0.0011644868449022885, "loss": 1.5176, "step": 4707 }, { "epoch": 0.46, "learning_rate": 0.0011641718856242645, "loss": 1.583, "step": 4708 }, { "epoch": 0.46, "learning_rate": 0.0011638569096085083, "loss": 1.5537, "step": 4709 }, { "epoch": 0.46, "learning_rate": 0.001163541916887133, "loss": 1.5986, "step": 4710 }, { "epoch": 0.46, "learning_rate": 0.0011632269074922528, "loss": 1.5195, "step": 4711 }, { "epoch": 0.46, "learning_rate": 0.0011629118814559835, "loss": 1.542, "step": 4712 }, { "epoch": 0.46, "learning_rate": 0.0011625968388104429, "loss": 1.6006, "step": 4713 }, { "epoch": 0.46, "learning_rate": 0.00116228177958775, "loss": 1.583, "step": 4714 }, { "epoch": 0.46, "learning_rate": 0.0011619667038200265, "loss": 1.5137, "step": 4715 }, { "epoch": 0.46, "learning_rate": 0.0011616516115393952, "loss": 1.5371, "step": 4716 }, { "epoch": 0.47, "learning_rate": 0.0011613365027779796, "loss": 1.6084, "step": 4717 }, { "epoch": 0.47, "learning_rate": 0.0011610213775679068, "loss": 1.5703, "step": 4718 }, { "epoch": 0.47, "learning_rate": 0.0011607062359413043, "loss": 1.666, "step": 4719 }, { "epoch": 0.47, "learning_rate": 0.0011603910779303015, "loss": 1.6406, "step": 4720 }, { "epoch": 0.47, "learning_rate": 0.0011600759035670294, "loss": 1.5146, "step": 4721 }, { "epoch": 0.47, "learning_rate": 0.001159760712883621, "loss": 1.5586, "step": 4722 }, { "epoch": 0.47, "learning_rate": 0.0011594455059122106, "loss": 1.6006, "step": 4723 }, { "epoch": 0.47, "learning_rate": 0.0011591302826849347, "loss": 1.6064, "step": 4724 }, { "epoch": 0.47, "learning_rate": 0.0011588150432339306, "loss": 1.6143, "step": 4725 }, { "epoch": 0.47, "learning_rate": 0.0011584997875913384, "loss": 1.5293, "step": 4726 }, { "epoch": 0.47, "learning_rate": 0.0011581845157892985, "loss": 1.5381, "step": 4727 }, { "epoch": 0.47, "learning_rate": 0.0011578692278599542, "loss": 1.5332, "step": 4728 }, { "epoch": 0.47, "learning_rate": 0.0011575539238354493, "loss": 1.5264, "step": 4729 }, { "epoch": 0.47, "learning_rate": 0.0011572386037479302, "loss": 1.5781, "step": 4730 }, { "epoch": 0.47, "learning_rate": 0.0011569232676295449, "loss": 1.5156, "step": 4731 }, { "epoch": 0.47, "learning_rate": 0.0011566079155124417, "loss": 1.5215, "step": 4732 }, { "epoch": 0.47, "learning_rate": 0.001156292547428773, "loss": 1.583, "step": 4733 }, { "epoch": 0.47, "learning_rate": 0.0011559771634106902, "loss": 1.6455, "step": 4734 }, { "epoch": 0.47, "learning_rate": 0.0011556617634903477, "loss": 1.5225, "step": 4735 }, { "epoch": 0.47, "learning_rate": 0.0011553463476999014, "loss": 1.5596, "step": 4736 }, { "epoch": 0.47, "learning_rate": 0.0011550309160715088, "loss": 1.5703, "step": 4737 }, { "epoch": 0.47, "learning_rate": 0.001154715468637329, "loss": 1.5752, "step": 4738 }, { "epoch": 0.47, "learning_rate": 0.0011544000054295222, "loss": 1.5273, "step": 4739 }, { "epoch": 0.47, "learning_rate": 0.001154084526480251, "loss": 1.5176, "step": 4740 }, { "epoch": 0.47, "learning_rate": 0.0011537690318216796, "loss": 1.5303, "step": 4741 }, { "epoch": 0.47, "learning_rate": 0.001153453521485973, "loss": 1.5996, "step": 4742 }, { "epoch": 0.47, "learning_rate": 0.0011531379955052982, "loss": 1.627, "step": 4743 }, { "epoch": 0.47, "learning_rate": 0.001152822453911824, "loss": 1.5752, "step": 4744 }, { "epoch": 0.47, "learning_rate": 0.0011525068967377212, "loss": 1.5752, "step": 4745 }, { "epoch": 0.47, "learning_rate": 0.0011521913240151607, "loss": 1.6172, "step": 4746 }, { "epoch": 0.47, "learning_rate": 0.0011518757357763165, "loss": 1.5254, "step": 4747 }, { "epoch": 0.47, "learning_rate": 0.0011515601320533633, "loss": 1.5234, "step": 4748 }, { "epoch": 0.47, "learning_rate": 0.0011512445128784778, "loss": 1.5713, "step": 4749 }, { "epoch": 0.47, "learning_rate": 0.0011509288782838385, "loss": 1.5332, "step": 4750 }, { "epoch": 0.47, "learning_rate": 0.0011506132283016246, "loss": 1.6035, "step": 4751 }, { "epoch": 0.47, "learning_rate": 0.0011502975629640178, "loss": 1.6377, "step": 4752 }, { "epoch": 0.47, "learning_rate": 0.0011499818823032006, "loss": 1.6299, "step": 4753 }, { "epoch": 0.47, "learning_rate": 0.0011496661863513579, "loss": 1.543, "step": 4754 }, { "epoch": 0.47, "learning_rate": 0.0011493504751406752, "loss": 1.6738, "step": 4755 }, { "epoch": 0.47, "learning_rate": 0.0011490347487033402, "loss": 1.5918, "step": 4756 }, { "epoch": 0.47, "learning_rate": 0.0011487190070715418, "loss": 1.585, "step": 4757 }, { "epoch": 0.47, "learning_rate": 0.0011484032502774716, "loss": 1.5186, "step": 4758 }, { "epoch": 0.47, "learning_rate": 0.0011480874783533206, "loss": 1.6396, "step": 4759 }, { "epoch": 0.47, "learning_rate": 0.001147771691331283, "loss": 1.5469, "step": 4760 }, { "epoch": 0.47, "learning_rate": 0.0011474558892435548, "loss": 1.5791, "step": 4761 }, { "epoch": 0.47, "learning_rate": 0.0011471400721223314, "loss": 1.6152, "step": 4762 }, { "epoch": 0.47, "learning_rate": 0.001146824239999812, "loss": 1.5215, "step": 4763 }, { "epoch": 0.47, "learning_rate": 0.001146508392908196, "loss": 1.5811, "step": 4764 }, { "epoch": 0.47, "learning_rate": 0.0011461925308796854, "loss": 1.459, "step": 4765 }, { "epoch": 0.47, "learning_rate": 0.0011458766539464828, "loss": 1.7334, "step": 4766 }, { "epoch": 0.47, "learning_rate": 0.0011455607621407926, "loss": 1.6338, "step": 4767 }, { "epoch": 0.47, "learning_rate": 0.001145244855494821, "loss": 1.4639, "step": 4768 }, { "epoch": 0.47, "learning_rate": 0.0011449289340407752, "loss": 1.5957, "step": 4769 }, { "epoch": 0.47, "learning_rate": 0.0011446129978108642, "loss": 1.6074, "step": 4770 }, { "epoch": 0.47, "learning_rate": 0.0011442970468372986, "loss": 1.5459, "step": 4771 }, { "epoch": 0.47, "learning_rate": 0.0011439810811522905, "loss": 1.4209, "step": 4772 }, { "epoch": 0.47, "learning_rate": 0.0011436651007880527, "loss": 1.543, "step": 4773 }, { "epoch": 0.47, "learning_rate": 0.001143349105776801, "loss": 1.5566, "step": 4774 }, { "epoch": 0.47, "learning_rate": 0.0011430330961507518, "loss": 1.5693, "step": 4775 }, { "epoch": 0.47, "learning_rate": 0.0011427170719421227, "loss": 1.5684, "step": 4776 }, { "epoch": 0.47, "learning_rate": 0.0011424010331831332, "loss": 1.6621, "step": 4777 }, { "epoch": 0.47, "learning_rate": 0.0011420849799060045, "loss": 1.5762, "step": 4778 }, { "epoch": 0.47, "learning_rate": 0.0011417689121429589, "loss": 1.4639, "step": 4779 }, { "epoch": 0.47, "learning_rate": 0.00114145282992622, "loss": 1.6514, "step": 4780 }, { "epoch": 0.47, "learning_rate": 0.0011411367332880137, "loss": 1.5547, "step": 4781 }, { "epoch": 0.47, "learning_rate": 0.001140820622260566, "loss": 1.6309, "step": 4782 }, { "epoch": 0.47, "learning_rate": 0.0011405044968761063, "loss": 1.7119, "step": 4783 }, { "epoch": 0.47, "learning_rate": 0.0011401883571668639, "loss": 1.5693, "step": 4784 }, { "epoch": 0.47, "learning_rate": 0.0011398722031650697, "loss": 1.625, "step": 4785 }, { "epoch": 0.47, "learning_rate": 0.0011395560349029567, "loss": 1.5371, "step": 4786 }, { "epoch": 0.47, "learning_rate": 0.0011392398524127589, "loss": 1.5791, "step": 4787 }, { "epoch": 0.47, "learning_rate": 0.001138923655726712, "loss": 1.543, "step": 4788 }, { "epoch": 0.47, "learning_rate": 0.0011386074448770536, "loss": 1.5576, "step": 4789 }, { "epoch": 0.47, "learning_rate": 0.001138291219896021, "loss": 1.5361, "step": 4790 }, { "epoch": 0.47, "learning_rate": 0.0011379749808158547, "loss": 1.5938, "step": 4791 }, { "epoch": 0.47, "learning_rate": 0.0011376587276687963, "loss": 1.5986, "step": 4792 }, { "epoch": 0.47, "learning_rate": 0.0011373424604870883, "loss": 1.5645, "step": 4793 }, { "epoch": 0.47, "learning_rate": 0.001137026179302975, "loss": 1.583, "step": 4794 }, { "epoch": 0.47, "learning_rate": 0.001136709884148702, "loss": 1.5918, "step": 4795 }, { "epoch": 0.47, "learning_rate": 0.0011363935750565164, "loss": 1.6445, "step": 4796 }, { "epoch": 0.47, "learning_rate": 0.0011360772520586668, "loss": 1.5205, "step": 4797 }, { "epoch": 0.47, "learning_rate": 0.001135760915187403, "loss": 1.5811, "step": 4798 }, { "epoch": 0.47, "learning_rate": 0.0011354445644749765, "loss": 1.5771, "step": 4799 }, { "epoch": 0.47, "learning_rate": 0.0011351281999536398, "loss": 1.5234, "step": 4800 }, { "epoch": 0.47, "learning_rate": 0.0011348118216556473, "loss": 1.4766, "step": 4801 }, { "epoch": 0.47, "learning_rate": 0.0011344954296132544, "loss": 1.5225, "step": 4802 }, { "epoch": 0.47, "learning_rate": 0.0011341790238587179, "loss": 1.6543, "step": 4803 }, { "epoch": 0.47, "learning_rate": 0.0011338626044242967, "loss": 1.582, "step": 4804 }, { "epoch": 0.47, "learning_rate": 0.0011335461713422499, "loss": 1.5186, "step": 4805 }, { "epoch": 0.47, "learning_rate": 0.0011332297246448388, "loss": 1.5938, "step": 4806 }, { "epoch": 0.47, "learning_rate": 0.0011329132643643264, "loss": 1.5908, "step": 4807 }, { "epoch": 0.47, "learning_rate": 0.001132596790532976, "loss": 1.5605, "step": 4808 }, { "epoch": 0.47, "learning_rate": 0.0011322803031830531, "loss": 1.5938, "step": 4809 }, { "epoch": 0.47, "learning_rate": 0.0011319638023468245, "loss": 1.6104, "step": 4810 }, { "epoch": 0.47, "learning_rate": 0.0011316472880565582, "loss": 1.6484, "step": 4811 }, { "epoch": 0.47, "learning_rate": 0.0011313307603445237, "loss": 1.6338, "step": 4812 }, { "epoch": 0.47, "learning_rate": 0.0011310142192429914, "loss": 1.6689, "step": 4813 }, { "epoch": 0.47, "learning_rate": 0.0011306976647842341, "loss": 1.4844, "step": 4814 }, { "epoch": 0.47, "learning_rate": 0.0011303810970005248, "loss": 1.4229, "step": 4815 }, { "epoch": 0.47, "learning_rate": 0.0011300645159241385, "loss": 1.6221, "step": 4816 }, { "epoch": 0.47, "learning_rate": 0.0011297479215873513, "loss": 1.6113, "step": 4817 }, { "epoch": 0.47, "learning_rate": 0.0011294313140224413, "loss": 1.5459, "step": 4818 }, { "epoch": 0.48, "learning_rate": 0.0011291146932616866, "loss": 1.5811, "step": 4819 }, { "epoch": 0.48, "learning_rate": 0.0011287980593373684, "loss": 1.6055, "step": 4820 }, { "epoch": 0.48, "learning_rate": 0.0011284814122817675, "loss": 1.4893, "step": 4821 }, { "epoch": 0.48, "learning_rate": 0.0011281647521271675, "loss": 1.5576, "step": 4822 }, { "epoch": 0.48, "learning_rate": 0.001127848078905852, "loss": 1.6289, "step": 4823 }, { "epoch": 0.48, "learning_rate": 0.001127531392650107, "loss": 1.6006, "step": 4824 }, { "epoch": 0.48, "learning_rate": 0.0011272146933922196, "loss": 1.5664, "step": 4825 }, { "epoch": 0.48, "learning_rate": 0.001126897981164478, "loss": 1.5215, "step": 4826 }, { "epoch": 0.48, "learning_rate": 0.0011265812559991717, "loss": 1.499, "step": 4827 }, { "epoch": 0.48, "learning_rate": 0.0011262645179285918, "loss": 1.5869, "step": 4828 }, { "epoch": 0.48, "learning_rate": 0.0011259477669850301, "loss": 1.4688, "step": 4829 }, { "epoch": 0.48, "learning_rate": 0.0011256310032007808, "loss": 1.5557, "step": 4830 }, { "epoch": 0.48, "learning_rate": 0.0011253142266081382, "loss": 1.5645, "step": 4831 }, { "epoch": 0.48, "learning_rate": 0.0011249974372393985, "loss": 1.4609, "step": 4832 }, { "epoch": 0.48, "learning_rate": 0.0011246806351268592, "loss": 1.5342, "step": 4833 }, { "epoch": 0.48, "learning_rate": 0.0011243638203028196, "loss": 1.5879, "step": 4834 }, { "epoch": 0.48, "learning_rate": 0.0011240469927995792, "loss": 1.6152, "step": 4835 }, { "epoch": 0.48, "learning_rate": 0.0011237301526494394, "loss": 1.4727, "step": 4836 }, { "epoch": 0.48, "learning_rate": 0.0011234132998847028, "loss": 1.4648, "step": 4837 }, { "epoch": 0.48, "learning_rate": 0.0011230964345376737, "loss": 1.5537, "step": 4838 }, { "epoch": 0.48, "learning_rate": 0.0011227795566406571, "loss": 1.5127, "step": 4839 }, { "epoch": 0.48, "learning_rate": 0.0011224626662259595, "loss": 1.626, "step": 4840 }, { "epoch": 0.48, "learning_rate": 0.0011221457633258886, "loss": 1.6113, "step": 4841 }, { "epoch": 0.48, "learning_rate": 0.001121828847972753, "loss": 1.5381, "step": 4842 }, { "epoch": 0.48, "learning_rate": 0.0011215119201988637, "loss": 1.5479, "step": 4843 }, { "epoch": 0.48, "learning_rate": 0.0011211949800365322, "loss": 1.6816, "step": 4844 }, { "epoch": 0.48, "learning_rate": 0.0011208780275180714, "loss": 1.5146, "step": 4845 }, { "epoch": 0.48, "learning_rate": 0.0011205610626757948, "loss": 1.5059, "step": 4846 }, { "epoch": 0.48, "learning_rate": 0.0011202440855420184, "loss": 1.5762, "step": 4847 }, { "epoch": 0.48, "learning_rate": 0.0011199270961490583, "loss": 1.5684, "step": 4848 }, { "epoch": 0.48, "learning_rate": 0.0011196100945292324, "loss": 1.5664, "step": 4849 }, { "epoch": 0.48, "learning_rate": 0.0011192930807148603, "loss": 1.5771, "step": 4850 }, { "epoch": 0.48, "learning_rate": 0.0011189760547382621, "loss": 1.6045, "step": 4851 }, { "epoch": 0.48, "learning_rate": 0.0011186590166317593, "loss": 1.6055, "step": 4852 }, { "epoch": 0.48, "learning_rate": 0.0011183419664276746, "loss": 1.5312, "step": 4853 }, { "epoch": 0.48, "learning_rate": 0.0011180249041583324, "loss": 1.5537, "step": 4854 }, { "epoch": 0.48, "learning_rate": 0.0011177078298560577, "loss": 1.626, "step": 4855 }, { "epoch": 0.48, "learning_rate": 0.0011173907435531772, "loss": 1.5869, "step": 4856 }, { "epoch": 0.48, "learning_rate": 0.0011170736452820183, "loss": 1.5771, "step": 4857 }, { "epoch": 0.48, "learning_rate": 0.0011167565350749102, "loss": 1.625, "step": 4858 }, { "epoch": 0.48, "learning_rate": 0.0011164394129641834, "loss": 1.417, "step": 4859 }, { "epoch": 0.48, "learning_rate": 0.0011161222789821685, "loss": 1.4795, "step": 4860 }, { "epoch": 0.48, "learning_rate": 0.001115805133161199, "loss": 1.582, "step": 4861 }, { "epoch": 0.48, "learning_rate": 0.0011154879755336082, "loss": 1.5137, "step": 4862 }, { "epoch": 0.48, "learning_rate": 0.0011151708061317312, "loss": 1.6191, "step": 4863 }, { "epoch": 0.48, "learning_rate": 0.001114853624987904, "loss": 1.6543, "step": 4864 }, { "epoch": 0.48, "learning_rate": 0.0011145364321344645, "loss": 1.5684, "step": 4865 }, { "epoch": 0.48, "learning_rate": 0.0011142192276037506, "loss": 1.5615, "step": 4866 }, { "epoch": 0.48, "learning_rate": 0.0011139020114281023, "loss": 1.6455, "step": 4867 }, { "epoch": 0.48, "learning_rate": 0.0011135847836398615, "loss": 1.5205, "step": 4868 }, { "epoch": 0.48, "learning_rate": 0.0011132675442713691, "loss": 1.5098, "step": 4869 }, { "epoch": 0.48, "learning_rate": 0.0011129502933549691, "loss": 1.6182, "step": 4870 }, { "epoch": 0.48, "learning_rate": 0.001112633030923006, "loss": 1.6748, "step": 4871 }, { "epoch": 0.48, "learning_rate": 0.0011123157570078254, "loss": 1.6045, "step": 4872 }, { "epoch": 0.48, "learning_rate": 0.001111998471641774, "loss": 1.7031, "step": 4873 }, { "epoch": 0.48, "learning_rate": 0.0011116811748571997, "loss": 1.5391, "step": 4874 }, { "epoch": 0.48, "learning_rate": 0.0011113638666864524, "loss": 1.5996, "step": 4875 }, { "epoch": 0.48, "learning_rate": 0.001111046547161882, "loss": 1.6553, "step": 4876 }, { "epoch": 0.48, "learning_rate": 0.0011107292163158398, "loss": 1.7012, "step": 4877 }, { "epoch": 0.48, "learning_rate": 0.001110411874180679, "loss": 1.5049, "step": 4878 }, { "epoch": 0.48, "learning_rate": 0.0011100945207887529, "loss": 1.6562, "step": 4879 }, { "epoch": 0.48, "learning_rate": 0.0011097771561724167, "loss": 1.4961, "step": 4880 }, { "epoch": 0.48, "learning_rate": 0.0011094597803640268, "loss": 1.4736, "step": 4881 }, { "epoch": 0.48, "learning_rate": 0.00110914239339594, "loss": 1.4766, "step": 4882 }, { "epoch": 0.48, "learning_rate": 0.0011088249953005145, "loss": 1.5918, "step": 4883 }, { "epoch": 0.48, "learning_rate": 0.0011085075861101104, "loss": 1.625, "step": 4884 }, { "epoch": 0.48, "learning_rate": 0.0011081901658570883, "loss": 1.5479, "step": 4885 }, { "epoch": 0.48, "learning_rate": 0.0011078727345738097, "loss": 1.5898, "step": 4886 }, { "epoch": 0.48, "learning_rate": 0.0011075552922926377, "loss": 1.5791, "step": 4887 }, { "epoch": 0.48, "learning_rate": 0.001107237839045936, "loss": 1.4668, "step": 4888 }, { "epoch": 0.48, "learning_rate": 0.0011069203748660704, "loss": 1.4521, "step": 4889 }, { "epoch": 0.48, "learning_rate": 0.0011066028997854065, "loss": 1.6719, "step": 4890 }, { "epoch": 0.48, "learning_rate": 0.001106285413836312, "loss": 1.7139, "step": 4891 }, { "epoch": 0.48, "learning_rate": 0.001105967917051155, "loss": 1.4727, "step": 4892 }, { "epoch": 0.48, "learning_rate": 0.001105650409462306, "loss": 1.6582, "step": 4893 }, { "epoch": 0.48, "learning_rate": 0.0011053328911021347, "loss": 1.6797, "step": 4894 }, { "epoch": 0.48, "learning_rate": 0.0011050153620030135, "loss": 1.5781, "step": 4895 }, { "epoch": 0.48, "learning_rate": 0.001104697822197315, "loss": 1.5283, "step": 4896 }, { "epoch": 0.48, "learning_rate": 0.0011043802717174132, "loss": 1.5908, "step": 4897 }, { "epoch": 0.48, "learning_rate": 0.0011040627105956833, "loss": 1.5117, "step": 4898 }, { "epoch": 0.48, "learning_rate": 0.0011037451388645012, "loss": 1.5479, "step": 4899 }, { "epoch": 0.48, "learning_rate": 0.0011034275565562444, "loss": 1.5732, "step": 4900 }, { "epoch": 0.48, "learning_rate": 0.001103109963703291, "loss": 1.6133, "step": 4901 }, { "epoch": 0.48, "learning_rate": 0.0011027923603380203, "loss": 1.6123, "step": 4902 }, { "epoch": 0.48, "learning_rate": 0.0011024747464928133, "loss": 1.6982, "step": 4903 }, { "epoch": 0.48, "learning_rate": 0.0011021571222000507, "loss": 1.5479, "step": 4904 }, { "epoch": 0.48, "learning_rate": 0.0011018394874921158, "loss": 1.5537, "step": 4905 }, { "epoch": 0.48, "learning_rate": 0.001101521842401392, "loss": 1.4268, "step": 4906 }, { "epoch": 0.48, "learning_rate": 0.0011012041869602638, "loss": 1.4951, "step": 4907 }, { "epoch": 0.48, "learning_rate": 0.001100886521201117, "loss": 1.4346, "step": 4908 }, { "epoch": 0.48, "learning_rate": 0.0011005688451563387, "loss": 1.5469, "step": 4909 }, { "epoch": 0.48, "learning_rate": 0.0011002511588583167, "loss": 1.4756, "step": 4910 }, { "epoch": 0.48, "learning_rate": 0.0010999334623394397, "loss": 1.5508, "step": 4911 }, { "epoch": 0.48, "learning_rate": 0.0010996157556320979, "loss": 1.6436, "step": 4912 }, { "epoch": 0.48, "learning_rate": 0.001099298038768682, "loss": 1.5508, "step": 4913 }, { "epoch": 0.48, "learning_rate": 0.0010989803117815843, "loss": 1.4893, "step": 4914 }, { "epoch": 0.48, "learning_rate": 0.0010986625747031978, "loss": 1.5615, "step": 4915 }, { "epoch": 0.48, "learning_rate": 0.0010983448275659162, "loss": 1.4912, "step": 4916 }, { "epoch": 0.48, "learning_rate": 0.0010980270704021352, "loss": 1.5693, "step": 4917 }, { "epoch": 0.48, "learning_rate": 0.0010977093032442503, "loss": 1.6465, "step": 4918 }, { "epoch": 0.48, "learning_rate": 0.0010973915261246595, "loss": 1.6016, "step": 4919 }, { "epoch": 0.49, "learning_rate": 0.0010970737390757603, "loss": 1.5518, "step": 4920 }, { "epoch": 0.49, "learning_rate": 0.0010967559421299522, "loss": 1.542, "step": 4921 }, { "epoch": 0.49, "learning_rate": 0.0010964381353196352, "loss": 1.5625, "step": 4922 }, { "epoch": 0.49, "learning_rate": 0.0010961203186772104, "loss": 1.6416, "step": 4923 }, { "epoch": 0.49, "learning_rate": 0.0010958024922350803, "loss": 1.5254, "step": 4924 }, { "epoch": 0.49, "learning_rate": 0.001095484656025648, "loss": 1.4062, "step": 4925 }, { "epoch": 0.49, "learning_rate": 0.0010951668100813172, "loss": 1.4307, "step": 4926 }, { "epoch": 0.49, "learning_rate": 0.0010948489544344937, "loss": 1.5518, "step": 4927 }, { "epoch": 0.49, "learning_rate": 0.0010945310891175837, "loss": 1.5586, "step": 4928 }, { "epoch": 0.49, "learning_rate": 0.0010942132141629941, "loss": 1.5469, "step": 4929 }, { "epoch": 0.49, "learning_rate": 0.0010938953296031333, "loss": 1.5957, "step": 4930 }, { "epoch": 0.49, "learning_rate": 0.0010935774354704098, "loss": 1.5674, "step": 4931 }, { "epoch": 0.49, "learning_rate": 0.0010932595317972346, "loss": 1.5977, "step": 4932 }, { "epoch": 0.49, "learning_rate": 0.0010929416186160174, "loss": 1.5996, "step": 4933 }, { "epoch": 0.49, "learning_rate": 0.0010926236959591718, "loss": 1.5605, "step": 4934 }, { "epoch": 0.49, "learning_rate": 0.00109230576385911, "loss": 1.5527, "step": 4935 }, { "epoch": 0.49, "learning_rate": 0.001091987822348246, "loss": 1.5957, "step": 4936 }, { "epoch": 0.49, "learning_rate": 0.001091669871458995, "loss": 1.4639, "step": 4937 }, { "epoch": 0.49, "learning_rate": 0.0010913519112237726, "loss": 1.5605, "step": 4938 }, { "epoch": 0.49, "learning_rate": 0.0010910339416749957, "loss": 1.6113, "step": 4939 }, { "epoch": 0.49, "learning_rate": 0.0010907159628450822, "loss": 1.5801, "step": 4940 }, { "epoch": 0.49, "learning_rate": 0.0010903979747664508, "loss": 1.5293, "step": 4941 }, { "epoch": 0.49, "learning_rate": 0.0010900799774715207, "loss": 1.6357, "step": 4942 }, { "epoch": 0.49, "learning_rate": 0.0010897619709927134, "loss": 1.541, "step": 4943 }, { "epoch": 0.49, "learning_rate": 0.0010894439553624498, "loss": 1.5596, "step": 4944 }, { "epoch": 0.49, "learning_rate": 0.0010891259306131528, "loss": 1.5479, "step": 4945 }, { "epoch": 0.49, "learning_rate": 0.0010888078967772454, "loss": 1.501, "step": 4946 }, { "epoch": 0.49, "learning_rate": 0.0010884898538871522, "loss": 1.5303, "step": 4947 }, { "epoch": 0.49, "learning_rate": 0.0010881718019752984, "loss": 1.4814, "step": 4948 }, { "epoch": 0.49, "learning_rate": 0.0010878537410741102, "loss": 1.5293, "step": 4949 }, { "epoch": 0.49, "learning_rate": 0.0010875356712160148, "loss": 1.5488, "step": 4950 }, { "epoch": 0.49, "learning_rate": 0.0010872175924334397, "loss": 1.6914, "step": 4951 }, { "epoch": 0.49, "learning_rate": 0.0010868995047588145, "loss": 1.6523, "step": 4952 }, { "epoch": 0.49, "learning_rate": 0.0010865814082245688, "loss": 1.5635, "step": 4953 }, { "epoch": 0.49, "learning_rate": 0.0010862633028631333, "loss": 1.6611, "step": 4954 }, { "epoch": 0.49, "learning_rate": 0.0010859451887069395, "loss": 1.6816, "step": 4955 }, { "epoch": 0.49, "learning_rate": 0.0010856270657884199, "loss": 1.5605, "step": 4956 }, { "epoch": 0.49, "learning_rate": 0.0010853089341400085, "loss": 1.4082, "step": 4957 }, { "epoch": 0.49, "learning_rate": 0.0010849907937941388, "loss": 1.6084, "step": 4958 }, { "epoch": 0.49, "learning_rate": 0.0010846726447832464, "loss": 1.5176, "step": 4959 }, { "epoch": 0.49, "learning_rate": 0.0010843544871397674, "loss": 1.5605, "step": 4960 }, { "epoch": 0.49, "learning_rate": 0.001084036320896139, "loss": 1.5693, "step": 4961 }, { "epoch": 0.49, "learning_rate": 0.0010837181460847986, "loss": 1.5537, "step": 4962 }, { "epoch": 0.49, "learning_rate": 0.0010833999627381848, "loss": 1.6172, "step": 4963 }, { "epoch": 0.49, "learning_rate": 0.0010830817708887377, "loss": 1.6182, "step": 4964 }, { "epoch": 0.49, "learning_rate": 0.0010827635705688975, "loss": 1.3936, "step": 4965 }, { "epoch": 0.49, "learning_rate": 0.0010824453618111054, "loss": 1.5752, "step": 4966 }, { "epoch": 0.49, "learning_rate": 0.0010821271446478035, "loss": 1.6016, "step": 4967 }, { "epoch": 0.49, "learning_rate": 0.0010818089191114353, "loss": 1.5195, "step": 4968 }, { "epoch": 0.49, "learning_rate": 0.0010814906852344445, "loss": 1.5898, "step": 4969 }, { "epoch": 0.49, "learning_rate": 0.0010811724430492757, "loss": 1.4922, "step": 4970 }, { "epoch": 0.49, "learning_rate": 0.0010808541925883743, "loss": 1.6006, "step": 4971 }, { "epoch": 0.49, "learning_rate": 0.0010805359338841873, "loss": 1.6143, "step": 4972 }, { "epoch": 0.49, "learning_rate": 0.0010802176669691614, "loss": 1.6123, "step": 4973 }, { "epoch": 0.49, "learning_rate": 0.001079899391875745, "loss": 1.5039, "step": 4974 }, { "epoch": 0.49, "learning_rate": 0.001079581108636387, "loss": 1.5498, "step": 4975 }, { "epoch": 0.49, "learning_rate": 0.0010792628172835367, "loss": 1.6221, "step": 4976 }, { "epoch": 0.49, "learning_rate": 0.0010789445178496456, "loss": 1.4385, "step": 4977 }, { "epoch": 0.49, "learning_rate": 0.0010786262103671645, "loss": 1.6426, "step": 4978 }, { "epoch": 0.49, "learning_rate": 0.001078307894868546, "loss": 1.6621, "step": 4979 }, { "epoch": 0.49, "learning_rate": 0.0010779895713862427, "loss": 1.5303, "step": 4980 }, { "epoch": 0.49, "learning_rate": 0.0010776712399527086, "loss": 1.583, "step": 4981 }, { "epoch": 0.49, "learning_rate": 0.0010773529006003987, "loss": 1.5918, "step": 4982 }, { "epoch": 0.49, "learning_rate": 0.0010770345533617682, "loss": 1.4795, "step": 4983 }, { "epoch": 0.49, "learning_rate": 0.0010767161982692735, "loss": 1.6104, "step": 4984 }, { "epoch": 0.49, "learning_rate": 0.0010763978353553715, "loss": 1.5791, "step": 4985 }, { "epoch": 0.49, "learning_rate": 0.0010760794646525204, "loss": 1.6719, "step": 4986 }, { "epoch": 0.49, "learning_rate": 0.0010757610861931788, "loss": 1.5791, "step": 4987 }, { "epoch": 0.49, "learning_rate": 0.001075442700009806, "loss": 1.5176, "step": 4988 }, { "epoch": 0.49, "learning_rate": 0.0010751243061348623, "loss": 1.5078, "step": 4989 }, { "epoch": 0.49, "learning_rate": 0.001074805904600809, "loss": 1.5127, "step": 4990 }, { "epoch": 0.49, "learning_rate": 0.0010744874954401078, "loss": 1.6309, "step": 4991 }, { "epoch": 0.49, "learning_rate": 0.001074169078685221, "loss": 1.4561, "step": 4992 }, { "epoch": 0.49, "learning_rate": 0.0010738506543686121, "loss": 1.6133, "step": 4993 }, { "epoch": 0.49, "learning_rate": 0.0010735322225227458, "loss": 1.5605, "step": 4994 }, { "epoch": 0.49, "learning_rate": 0.0010732137831800863, "loss": 1.582, "step": 4995 }, { "epoch": 0.49, "learning_rate": 0.0010728953363730999, "loss": 1.5674, "step": 4996 }, { "epoch": 0.49, "learning_rate": 0.0010725768821342526, "loss": 1.582, "step": 4997 }, { "epoch": 0.49, "learning_rate": 0.0010722584204960117, "loss": 1.5215, "step": 4998 }, { "epoch": 0.49, "learning_rate": 0.0010719399514908451, "loss": 1.5137, "step": 4999 }, { "epoch": 0.49, "learning_rate": 0.0010716214751512216, "loss": 1.5693, "step": 5000 }, { "epoch": 0.49, "learning_rate": 0.0010713029915096106, "loss": 1.5762, "step": 5001 }, { "epoch": 0.49, "learning_rate": 0.0010709845005984827, "loss": 1.5264, "step": 5002 }, { "epoch": 0.49, "learning_rate": 0.001070666002450308, "loss": 1.5508, "step": 5003 }, { "epoch": 0.49, "learning_rate": 0.001070347497097559, "loss": 1.6016, "step": 5004 }, { "epoch": 0.49, "learning_rate": 0.0010700289845727079, "loss": 1.4824, "step": 5005 }, { "epoch": 0.49, "learning_rate": 0.0010697104649082273, "loss": 1.6709, "step": 5006 }, { "epoch": 0.49, "learning_rate": 0.0010693919381365917, "loss": 1.6826, "step": 5007 }, { "epoch": 0.49, "learning_rate": 0.0010690734042902758, "loss": 1.5928, "step": 5008 }, { "epoch": 0.49, "learning_rate": 0.001068754863401754, "loss": 1.5801, "step": 5009 }, { "epoch": 0.49, "learning_rate": 0.0010684363155035032, "loss": 1.5371, "step": 5010 }, { "epoch": 0.49, "learning_rate": 0.001068117760628, "loss": 1.5371, "step": 5011 }, { "epoch": 0.49, "learning_rate": 0.0010677991988077217, "loss": 1.5781, "step": 5012 }, { "epoch": 0.49, "learning_rate": 0.0010674806300751466, "loss": 1.5107, "step": 5013 }, { "epoch": 0.49, "learning_rate": 0.0010671620544627535, "loss": 1.4775, "step": 5014 }, { "epoch": 0.49, "learning_rate": 0.001066843472003022, "loss": 1.5742, "step": 5015 }, { "epoch": 0.49, "learning_rate": 0.0010665248827284324, "loss": 1.5234, "step": 5016 }, { "epoch": 0.49, "learning_rate": 0.0010662062866714655, "loss": 1.4668, "step": 5017 }, { "epoch": 0.49, "learning_rate": 0.0010658876838646027, "loss": 1.6943, "step": 5018 }, { "epoch": 0.49, "learning_rate": 0.001065569074340327, "loss": 1.6074, "step": 5019 }, { "epoch": 0.49, "learning_rate": 0.0010652504581311214, "loss": 1.5801, "step": 5020 }, { "epoch": 0.49, "learning_rate": 0.0010649318352694693, "loss": 1.7227, "step": 5021 }, { "epoch": 0.5, "learning_rate": 0.0010646132057878551, "loss": 1.6846, "step": 5022 }, { "epoch": 0.5, "learning_rate": 0.0010642945697187638, "loss": 1.5078, "step": 5023 }, { "epoch": 0.5, "learning_rate": 0.0010639759270946814, "loss": 1.5186, "step": 5024 }, { "epoch": 0.5, "learning_rate": 0.0010636572779480942, "loss": 1.6025, "step": 5025 }, { "epoch": 0.5, "learning_rate": 0.001063338622311489, "loss": 1.543, "step": 5026 }, { "epoch": 0.5, "learning_rate": 0.001063019960217354, "loss": 1.7715, "step": 5027 }, { "epoch": 0.5, "learning_rate": 0.0010627012916981772, "loss": 1.4482, "step": 5028 }, { "epoch": 0.5, "learning_rate": 0.001062382616786448, "loss": 1.6104, "step": 5029 }, { "epoch": 0.5, "learning_rate": 0.0010620639355146562, "loss": 1.4893, "step": 5030 }, { "epoch": 0.5, "learning_rate": 0.0010617452479152916, "loss": 1.3867, "step": 5031 }, { "epoch": 0.5, "learning_rate": 0.0010614265540208453, "loss": 1.5557, "step": 5032 }, { "epoch": 0.5, "learning_rate": 0.0010611078538638095, "loss": 1.6934, "step": 5033 }, { "epoch": 0.5, "learning_rate": 0.0010607891474766758, "loss": 1.6523, "step": 5034 }, { "epoch": 0.5, "learning_rate": 0.0010604704348919373, "loss": 1.5469, "step": 5035 }, { "epoch": 0.5, "learning_rate": 0.0010601517161420878, "loss": 1.4824, "step": 5036 }, { "epoch": 0.5, "learning_rate": 0.0010598329912596212, "loss": 1.6611, "step": 5037 }, { "epoch": 0.5, "learning_rate": 0.0010595142602770325, "loss": 1.3955, "step": 5038 }, { "epoch": 0.5, "learning_rate": 0.0010591955232268169, "loss": 1.4717, "step": 5039 }, { "epoch": 0.5, "learning_rate": 0.0010588767801414708, "loss": 1.6426, "step": 5040 }, { "epoch": 0.5, "learning_rate": 0.0010585580310534903, "loss": 1.4824, "step": 5041 }, { "epoch": 0.5, "learning_rate": 0.0010582392759953731, "loss": 1.4385, "step": 5042 }, { "epoch": 0.5, "learning_rate": 0.0010579205149996167, "loss": 1.4326, "step": 5043 }, { "epoch": 0.5, "learning_rate": 0.0010576017480987202, "loss": 1.4717, "step": 5044 }, { "epoch": 0.5, "learning_rate": 0.001057282975325182, "loss": 1.582, "step": 5045 }, { "epoch": 0.5, "learning_rate": 0.001056964196711502, "loss": 1.4482, "step": 5046 }, { "epoch": 0.5, "learning_rate": 0.0010566454122901807, "loss": 1.5488, "step": 5047 }, { "epoch": 0.5, "learning_rate": 0.0010563266220937186, "loss": 1.6426, "step": 5048 }, { "epoch": 0.5, "learning_rate": 0.0010560078261546175, "loss": 1.6777, "step": 5049 }, { "epoch": 0.5, "learning_rate": 0.0010556890245053791, "loss": 1.6064, "step": 5050 }, { "epoch": 0.5, "learning_rate": 0.0010553702171785064, "loss": 1.5332, "step": 5051 }, { "epoch": 0.5, "learning_rate": 0.0010550514042065022, "loss": 1.582, "step": 5052 }, { "epoch": 0.5, "learning_rate": 0.0010547325856218705, "loss": 1.5635, "step": 5053 }, { "epoch": 0.5, "learning_rate": 0.0010544137614571158, "loss": 1.4902, "step": 5054 }, { "epoch": 0.5, "learning_rate": 0.001054094931744743, "loss": 1.5273, "step": 5055 }, { "epoch": 0.5, "learning_rate": 0.001053776096517257, "loss": 1.46, "step": 5056 }, { "epoch": 0.5, "learning_rate": 0.0010534572558071648, "loss": 1.6045, "step": 5057 }, { "epoch": 0.5, "learning_rate": 0.001053138409646972, "loss": 1.5889, "step": 5058 }, { "epoch": 0.5, "learning_rate": 0.0010528195580691863, "loss": 1.4131, "step": 5059 }, { "epoch": 0.5, "learning_rate": 0.0010525007011063154, "loss": 1.5918, "step": 5060 }, { "epoch": 0.5, "learning_rate": 0.0010521818387908672, "loss": 1.5762, "step": 5061 }, { "epoch": 0.5, "learning_rate": 0.0010518629711553514, "loss": 1.6416, "step": 5062 }, { "epoch": 0.5, "learning_rate": 0.0010515440982322764, "loss": 1.4561, "step": 5063 }, { "epoch": 0.5, "learning_rate": 0.0010512252200541522, "loss": 1.4453, "step": 5064 }, { "epoch": 0.5, "learning_rate": 0.0010509063366534901, "loss": 1.5898, "step": 5065 }, { "epoch": 0.5, "learning_rate": 0.0010505874480628, "loss": 1.5293, "step": 5066 }, { "epoch": 0.5, "learning_rate": 0.001050268554314594, "loss": 1.4941, "step": 5067 }, { "epoch": 0.5, "learning_rate": 0.0010499496554413836, "loss": 1.5459, "step": 5068 }, { "epoch": 0.5, "learning_rate": 0.001049630751475682, "loss": 1.4434, "step": 5069 }, { "epoch": 0.5, "learning_rate": 0.0010493118424500017, "loss": 1.6318, "step": 5070 }, { "epoch": 0.5, "learning_rate": 0.0010489929283968565, "loss": 1.6162, "step": 5071 }, { "epoch": 0.5, "learning_rate": 0.0010486740093487606, "loss": 1.3809, "step": 5072 }, { "epoch": 0.5, "learning_rate": 0.0010483550853382282, "loss": 1.6133, "step": 5073 }, { "epoch": 0.5, "learning_rate": 0.0010480361563977747, "loss": 1.584, "step": 5074 }, { "epoch": 0.5, "learning_rate": 0.0010477172225599157, "loss": 1.6084, "step": 5075 }, { "epoch": 0.5, "learning_rate": 0.001047398283857167, "loss": 1.46, "step": 5076 }, { "epoch": 0.5, "learning_rate": 0.0010470793403220455, "loss": 1.499, "step": 5077 }, { "epoch": 0.5, "learning_rate": 0.0010467603919870686, "loss": 1.5771, "step": 5078 }, { "epoch": 0.5, "learning_rate": 0.0010464414388847531, "loss": 1.5068, "step": 5079 }, { "epoch": 0.5, "learning_rate": 0.0010461224810476176, "loss": 1.5371, "step": 5080 }, { "epoch": 0.5, "learning_rate": 0.0010458035185081805, "loss": 1.5742, "step": 5081 }, { "epoch": 0.5, "learning_rate": 0.0010454845512989609, "loss": 1.6641, "step": 5082 }, { "epoch": 0.5, "learning_rate": 0.001045165579452478, "loss": 1.5781, "step": 5083 }, { "epoch": 0.5, "learning_rate": 0.0010448466030012521, "loss": 1.5156, "step": 5084 }, { "epoch": 0.5, "learning_rate": 0.0010445276219778032, "loss": 1.5127, "step": 5085 }, { "epoch": 0.5, "learning_rate": 0.0010442086364146528, "loss": 1.543, "step": 5086 }, { "epoch": 0.5, "learning_rate": 0.001043889646344322, "loss": 1.5498, "step": 5087 }, { "epoch": 0.5, "learning_rate": 0.0010435706517993326, "loss": 1.4727, "step": 5088 }, { "epoch": 0.5, "learning_rate": 0.001043251652812207, "loss": 1.627, "step": 5089 }, { "epoch": 0.5, "learning_rate": 0.0010429326494154681, "loss": 1.6279, "step": 5090 }, { "epoch": 0.5, "learning_rate": 0.0010426136416416384, "loss": 1.6172, "step": 5091 }, { "epoch": 0.5, "learning_rate": 0.0010422946295232426, "loss": 1.5635, "step": 5092 }, { "epoch": 0.5, "learning_rate": 0.0010419756130928036, "loss": 1.6689, "step": 5093 }, { "epoch": 0.5, "learning_rate": 0.0010416565923828467, "loss": 1.5244, "step": 5094 }, { "epoch": 0.5, "learning_rate": 0.0010413375674258965, "loss": 1.5811, "step": 5095 }, { "epoch": 0.5, "learning_rate": 0.0010410185382544786, "loss": 1.5098, "step": 5096 }, { "epoch": 0.5, "learning_rate": 0.0010406995049011187, "loss": 1.4395, "step": 5097 }, { "epoch": 0.5, "learning_rate": 0.0010403804673983434, "loss": 1.5322, "step": 5098 }, { "epoch": 0.5, "learning_rate": 0.001040061425778679, "loss": 1.6035, "step": 5099 }, { "epoch": 0.5, "learning_rate": 0.0010397423800746527, "loss": 1.5596, "step": 5100 }, { "epoch": 0.5, "learning_rate": 0.0010394233303187917, "loss": 1.458, "step": 5101 }, { "epoch": 0.5, "learning_rate": 0.0010391042765436238, "loss": 1.5771, "step": 5102 }, { "epoch": 0.5, "learning_rate": 0.0010387852187816783, "loss": 1.6289, "step": 5103 }, { "epoch": 0.5, "learning_rate": 0.001038466157065483, "loss": 1.4209, "step": 5104 }, { "epoch": 0.5, "learning_rate": 0.0010381470914275675, "loss": 1.4863, "step": 5105 }, { "epoch": 0.5, "learning_rate": 0.0010378280219004612, "loss": 1.5645, "step": 5106 }, { "epoch": 0.5, "learning_rate": 0.001037508948516694, "loss": 1.6016, "step": 5107 }, { "epoch": 0.5, "learning_rate": 0.0010371898713087962, "loss": 1.5361, "step": 5108 }, { "epoch": 0.5, "learning_rate": 0.0010368707903092988, "loss": 1.7158, "step": 5109 }, { "epoch": 0.5, "learning_rate": 0.0010365517055507323, "loss": 1.5977, "step": 5110 }, { "epoch": 0.5, "learning_rate": 0.0010362326170656284, "loss": 1.5488, "step": 5111 }, { "epoch": 0.5, "learning_rate": 0.0010359135248865195, "loss": 1.5381, "step": 5112 }, { "epoch": 0.5, "learning_rate": 0.0010355944290459374, "loss": 1.585, "step": 5113 }, { "epoch": 0.5, "learning_rate": 0.0010352753295764148, "loss": 1.6182, "step": 5114 }, { "epoch": 0.5, "learning_rate": 0.0010349562265104847, "loss": 1.5146, "step": 5115 }, { "epoch": 0.5, "learning_rate": 0.0010346371198806802, "loss": 1.5303, "step": 5116 }, { "epoch": 0.5, "learning_rate": 0.0010343180097195353, "loss": 1.542, "step": 5117 }, { "epoch": 0.5, "learning_rate": 0.001033998896059584, "loss": 1.5752, "step": 5118 }, { "epoch": 0.5, "learning_rate": 0.0010336797789333608, "loss": 1.6064, "step": 5119 }, { "epoch": 0.5, "learning_rate": 0.0010333606583734, "loss": 1.5723, "step": 5120 }, { "epoch": 0.5, "learning_rate": 0.0010330415344122375, "loss": 1.6426, "step": 5121 }, { "epoch": 0.5, "learning_rate": 0.0010327224070824084, "loss": 1.5752, "step": 5122 }, { "epoch": 0.51, "learning_rate": 0.0010324032764164484, "loss": 1.5801, "step": 5123 }, { "epoch": 0.51, "learning_rate": 0.001032084142446894, "loss": 1.541, "step": 5124 }, { "epoch": 0.51, "learning_rate": 0.0010317650052062814, "loss": 1.5898, "step": 5125 }, { "epoch": 0.51, "learning_rate": 0.0010314458647271477, "loss": 1.5615, "step": 5126 }, { "epoch": 0.51, "learning_rate": 0.00103112672104203, "loss": 1.5303, "step": 5127 }, { "epoch": 0.51, "learning_rate": 0.0010308075741834652, "loss": 1.6396, "step": 5128 }, { "epoch": 0.51, "learning_rate": 0.001030488424183992, "loss": 1.4473, "step": 5129 }, { "epoch": 0.51, "learning_rate": 0.0010301692710761481, "loss": 1.5771, "step": 5130 }, { "epoch": 0.51, "learning_rate": 0.0010298501148924722, "loss": 1.5859, "step": 5131 }, { "epoch": 0.51, "learning_rate": 0.0010295309556655025, "loss": 1.667, "step": 5132 }, { "epoch": 0.51, "learning_rate": 0.0010292117934277787, "loss": 1.4219, "step": 5133 }, { "epoch": 0.51, "learning_rate": 0.0010288926282118398, "loss": 1.5879, "step": 5134 }, { "epoch": 0.51, "learning_rate": 0.0010285734600502259, "loss": 1.5439, "step": 5135 }, { "epoch": 0.51, "learning_rate": 0.0010282542889754763, "loss": 1.4873, "step": 5136 }, { "epoch": 0.51, "learning_rate": 0.0010279351150201317, "loss": 1.5449, "step": 5137 }, { "epoch": 0.51, "learning_rate": 0.001027615938216733, "loss": 1.6387, "step": 5138 }, { "epoch": 0.51, "learning_rate": 0.0010272967585978207, "loss": 1.5176, "step": 5139 }, { "epoch": 0.51, "learning_rate": 0.001026977576195936, "loss": 1.6143, "step": 5140 }, { "epoch": 0.51, "learning_rate": 0.00102665839104362, "loss": 1.6328, "step": 5141 }, { "epoch": 0.51, "learning_rate": 0.0010263392031734152, "loss": 1.5576, "step": 5142 }, { "epoch": 0.51, "learning_rate": 0.001026020012617863, "loss": 1.5352, "step": 5143 }, { "epoch": 0.51, "learning_rate": 0.0010257008194095056, "loss": 1.457, "step": 5144 }, { "epoch": 0.51, "learning_rate": 0.0010253816235808856, "loss": 1.5742, "step": 5145 }, { "epoch": 0.51, "learning_rate": 0.0010250624251645466, "loss": 1.4912, "step": 5146 }, { "epoch": 0.51, "learning_rate": 0.0010247432241930307, "loss": 1.6367, "step": 5147 }, { "epoch": 0.51, "learning_rate": 0.0010244240206988818, "loss": 1.3877, "step": 5148 }, { "epoch": 0.51, "learning_rate": 0.001024104814714643, "loss": 1.582, "step": 5149 }, { "epoch": 0.51, "learning_rate": 0.0010237856062728585, "loss": 1.46, "step": 5150 }, { "epoch": 0.51, "learning_rate": 0.0010234663954060723, "loss": 1.6709, "step": 5151 }, { "epoch": 0.51, "learning_rate": 0.001023147182146829, "loss": 1.5273, "step": 5152 }, { "epoch": 0.51, "learning_rate": 0.0010228279665276728, "loss": 1.4971, "step": 5153 }, { "epoch": 0.51, "learning_rate": 0.0010225087485811481, "loss": 1.5615, "step": 5154 }, { "epoch": 0.51, "learning_rate": 0.0010221895283398013, "loss": 1.5225, "step": 5155 }, { "epoch": 0.51, "learning_rate": 0.0010218703058361767, "loss": 1.6162, "step": 5156 }, { "epoch": 0.51, "learning_rate": 0.0010215510811028199, "loss": 1.5811, "step": 5157 }, { "epoch": 0.51, "learning_rate": 0.001021231854172277, "loss": 1.5479, "step": 5158 }, { "epoch": 0.51, "learning_rate": 0.0010209126250770935, "loss": 1.5234, "step": 5159 }, { "epoch": 0.51, "learning_rate": 0.0010205933938498162, "loss": 1.5996, "step": 5160 }, { "epoch": 0.51, "learning_rate": 0.0010202741605229908, "loss": 1.5273, "step": 5161 }, { "epoch": 0.51, "learning_rate": 0.0010199549251291645, "loss": 1.5986, "step": 5162 }, { "epoch": 0.51, "learning_rate": 0.0010196356877008842, "loss": 1.458, "step": 5163 }, { "epoch": 0.51, "learning_rate": 0.0010193164482706965, "loss": 1.6191, "step": 5164 }, { "epoch": 0.51, "learning_rate": 0.001018997206871149, "loss": 1.4824, "step": 5165 }, { "epoch": 0.51, "learning_rate": 0.001018677963534789, "loss": 1.4922, "step": 5166 }, { "epoch": 0.51, "learning_rate": 0.001018358718294164, "loss": 1.6445, "step": 5167 }, { "epoch": 0.51, "learning_rate": 0.0010180394711818222, "loss": 1.4639, "step": 5168 }, { "epoch": 0.51, "learning_rate": 0.0010177202222303114, "loss": 1.5078, "step": 5169 }, { "epoch": 0.51, "learning_rate": 0.0010174009714721797, "loss": 1.6611, "step": 5170 }, { "epoch": 0.51, "learning_rate": 0.0010170817189399759, "loss": 1.6279, "step": 5171 }, { "epoch": 0.51, "learning_rate": 0.0010167624646662485, "loss": 1.5488, "step": 5172 }, { "epoch": 0.51, "learning_rate": 0.0010164432086835463, "loss": 1.5205, "step": 5173 }, { "epoch": 0.51, "learning_rate": 0.0010161239510244177, "loss": 1.4854, "step": 5174 }, { "epoch": 0.51, "learning_rate": 0.0010158046917214123, "loss": 1.5518, "step": 5175 }, { "epoch": 0.51, "learning_rate": 0.0010154854308070792, "loss": 1.6016, "step": 5176 }, { "epoch": 0.51, "learning_rate": 0.0010151661683139683, "loss": 1.6572, "step": 5177 }, { "epoch": 0.51, "learning_rate": 0.0010148469042746287, "loss": 1.4746, "step": 5178 }, { "epoch": 0.51, "learning_rate": 0.00101452763872161, "loss": 1.5469, "step": 5179 }, { "epoch": 0.51, "learning_rate": 0.0010142083716874627, "loss": 1.5771, "step": 5180 }, { "epoch": 0.51, "learning_rate": 0.0010138891032047364, "loss": 1.541, "step": 5181 }, { "epoch": 0.51, "learning_rate": 0.0010135698333059817, "loss": 1.6895, "step": 5182 }, { "epoch": 0.51, "learning_rate": 0.001013250562023749, "loss": 1.4434, "step": 5183 }, { "epoch": 0.51, "learning_rate": 0.0010129312893905882, "loss": 1.6113, "step": 5184 }, { "epoch": 0.51, "learning_rate": 0.0010126120154390504, "loss": 1.5332, "step": 5185 }, { "epoch": 0.51, "learning_rate": 0.001012292740201686, "loss": 1.5713, "step": 5186 }, { "epoch": 0.51, "learning_rate": 0.0010119734637110463, "loss": 1.4932, "step": 5187 }, { "epoch": 0.51, "learning_rate": 0.001011654185999682, "loss": 1.5371, "step": 5188 }, { "epoch": 0.51, "learning_rate": 0.001011334907100145, "loss": 1.6201, "step": 5189 }, { "epoch": 0.51, "learning_rate": 0.0010110156270449857, "loss": 1.5342, "step": 5190 }, { "epoch": 0.51, "learning_rate": 0.001010696345866756, "loss": 1.5918, "step": 5191 }, { "epoch": 0.51, "learning_rate": 0.001010377063598007, "loss": 1.5459, "step": 5192 }, { "epoch": 0.51, "learning_rate": 0.001010057780271291, "loss": 1.5107, "step": 5193 }, { "epoch": 0.51, "learning_rate": 0.001009738495919159, "loss": 1.6631, "step": 5194 }, { "epoch": 0.51, "learning_rate": 0.0010094192105741628, "loss": 1.4033, "step": 5195 }, { "epoch": 0.51, "learning_rate": 0.0010090999242688553, "loss": 1.541, "step": 5196 }, { "epoch": 0.51, "learning_rate": 0.0010087806370357875, "loss": 1.5342, "step": 5197 }, { "epoch": 0.51, "learning_rate": 0.001008461348907512, "loss": 1.4941, "step": 5198 }, { "epoch": 0.51, "learning_rate": 0.001008142059916581, "loss": 1.5322, "step": 5199 }, { "epoch": 0.51, "learning_rate": 0.0010078227700955468, "loss": 1.6123, "step": 5200 }, { "epoch": 0.51, "learning_rate": 0.0010075034794769617, "loss": 1.5352, "step": 5201 }, { "epoch": 0.51, "learning_rate": 0.0010071841880933783, "loss": 1.4688, "step": 5202 }, { "epoch": 0.51, "learning_rate": 0.001006864895977349, "loss": 1.5791, "step": 5203 }, { "epoch": 0.51, "learning_rate": 0.0010065456031614262, "loss": 1.4512, "step": 5204 }, { "epoch": 0.51, "learning_rate": 0.0010062263096781631, "loss": 1.5439, "step": 5205 }, { "epoch": 0.51, "learning_rate": 0.0010059070155601121, "loss": 1.5381, "step": 5206 }, { "epoch": 0.51, "learning_rate": 0.0010055877208398265, "loss": 1.4482, "step": 5207 }, { "epoch": 0.51, "learning_rate": 0.0010052684255498587, "loss": 1.501, "step": 5208 }, { "epoch": 0.51, "learning_rate": 0.001004949129722762, "loss": 1.6465, "step": 5209 }, { "epoch": 0.51, "learning_rate": 0.0010046298333910892, "loss": 1.583, "step": 5210 }, { "epoch": 0.51, "learning_rate": 0.001004310536587393, "loss": 1.6055, "step": 5211 }, { "epoch": 0.51, "learning_rate": 0.0010039912393442272, "loss": 1.5918, "step": 5212 }, { "epoch": 0.51, "learning_rate": 0.0010036719416941446, "loss": 1.583, "step": 5213 }, { "epoch": 0.51, "learning_rate": 0.0010033526436696984, "loss": 1.5195, "step": 5214 }, { "epoch": 0.51, "learning_rate": 0.0010030333453034416, "loss": 1.6367, "step": 5215 }, { "epoch": 0.51, "learning_rate": 0.0010027140466279281, "loss": 1.5586, "step": 5216 }, { "epoch": 0.51, "learning_rate": 0.0010023947476757106, "loss": 1.5322, "step": 5217 }, { "epoch": 0.51, "learning_rate": 0.0010020754484793427, "loss": 1.5801, "step": 5218 }, { "epoch": 0.51, "learning_rate": 0.0010017561490713776, "loss": 1.6143, "step": 5219 }, { "epoch": 0.51, "learning_rate": 0.0010014368494843683, "loss": 1.5371, "step": 5220 }, { "epoch": 0.51, "learning_rate": 0.0010011175497508693, "loss": 1.5752, "step": 5221 }, { "epoch": 0.51, "learning_rate": 0.0010007982499034332, "loss": 1.501, "step": 5222 }, { "epoch": 0.51, "learning_rate": 0.0010004789499746133, "loss": 1.6602, "step": 5223 }, { "epoch": 0.51, "learning_rate": 0.0010001596499969634, "loss": 1.416, "step": 5224 }, { "epoch": 0.52, "learning_rate": 0.000999840350003037, "loss": 1.499, "step": 5225 }, { "epoch": 0.52, "learning_rate": 0.000999521050025387, "loss": 1.6113, "step": 5226 }, { "epoch": 0.52, "learning_rate": 0.0009992017500965672, "loss": 1.6182, "step": 5227 }, { "epoch": 0.52, "learning_rate": 0.000998882450249131, "loss": 1.4956, "step": 5228 }, { "epoch": 0.52, "learning_rate": 0.0009985631505156317, "loss": 1.6035, "step": 5229 }, { "epoch": 0.52, "learning_rate": 0.0009982438509286227, "loss": 1.4619, "step": 5230 }, { "epoch": 0.52, "learning_rate": 0.0009979245515206578, "loss": 1.5347, "step": 5231 }, { "epoch": 0.52, "learning_rate": 0.0009976052523242899, "loss": 1.502, "step": 5232 }, { "epoch": 0.52, "learning_rate": 0.0009972859533720724, "loss": 1.5205, "step": 5233 }, { "epoch": 0.52, "learning_rate": 0.0009969666546965586, "loss": 1.4326, "step": 5234 }, { "epoch": 0.52, "learning_rate": 0.000996647356330302, "loss": 1.3984, "step": 5235 }, { "epoch": 0.52, "learning_rate": 0.000996328058305856, "loss": 1.5186, "step": 5236 }, { "epoch": 0.52, "learning_rate": 0.000996008760655773, "loss": 1.459, "step": 5237 }, { "epoch": 0.52, "learning_rate": 0.000995689463412607, "loss": 1.6133, "step": 5238 }, { "epoch": 0.52, "learning_rate": 0.000995370166608911, "loss": 1.4717, "step": 5239 }, { "epoch": 0.52, "learning_rate": 0.0009950508702772382, "loss": 1.666, "step": 5240 }, { "epoch": 0.52, "learning_rate": 0.0009947315744501413, "loss": 1.6025, "step": 5241 }, { "epoch": 0.52, "learning_rate": 0.0009944122791601736, "loss": 1.4551, "step": 5242 }, { "epoch": 0.52, "learning_rate": 0.000994092984439888, "loss": 1.5615, "step": 5243 }, { "epoch": 0.52, "learning_rate": 0.0009937736903218371, "loss": 1.585, "step": 5244 }, { "epoch": 0.52, "learning_rate": 0.0009934543968385739, "loss": 1.6338, "step": 5245 }, { "epoch": 0.52, "learning_rate": 0.0009931351040226512, "loss": 1.3477, "step": 5246 }, { "epoch": 0.52, "learning_rate": 0.0009928158119066218, "loss": 1.4932, "step": 5247 }, { "epoch": 0.52, "learning_rate": 0.0009924965205230383, "loss": 1.585, "step": 5248 }, { "epoch": 0.52, "learning_rate": 0.0009921772299044532, "loss": 1.5615, "step": 5249 }, { "epoch": 0.52, "learning_rate": 0.000991857940083419, "loss": 1.4297, "step": 5250 }, { "epoch": 0.52, "learning_rate": 0.000991538651092488, "loss": 1.4438, "step": 5251 }, { "epoch": 0.52, "learning_rate": 0.0009912193629642127, "loss": 1.5615, "step": 5252 }, { "epoch": 0.52, "learning_rate": 0.000990900075731145, "loss": 1.5674, "step": 5253 }, { "epoch": 0.52, "learning_rate": 0.0009905807894258372, "loss": 1.5557, "step": 5254 }, { "epoch": 0.52, "learning_rate": 0.0009902615040808412, "loss": 1.4482, "step": 5255 }, { "epoch": 0.52, "learning_rate": 0.0009899422197287091, "loss": 1.5918, "step": 5256 }, { "epoch": 0.52, "learning_rate": 0.000989622936401993, "loss": 1.5898, "step": 5257 }, { "epoch": 0.52, "learning_rate": 0.000989303654133244, "loss": 1.6191, "step": 5258 }, { "epoch": 0.52, "learning_rate": 0.0009889843729550144, "loss": 1.5527, "step": 5259 }, { "epoch": 0.52, "learning_rate": 0.0009886650928998552, "loss": 1.5049, "step": 5260 }, { "epoch": 0.52, "learning_rate": 0.000988345814000318, "loss": 1.5508, "step": 5261 }, { "epoch": 0.52, "learning_rate": 0.000988026536288954, "loss": 1.5869, "step": 5262 }, { "epoch": 0.52, "learning_rate": 0.0009877072597983144, "loss": 1.708, "step": 5263 }, { "epoch": 0.52, "learning_rate": 0.0009873879845609499, "loss": 1.3945, "step": 5264 }, { "epoch": 0.52, "learning_rate": 0.000987068710609412, "loss": 1.5996, "step": 5265 }, { "epoch": 0.52, "learning_rate": 0.0009867494379762513, "loss": 1.5225, "step": 5266 }, { "epoch": 0.52, "learning_rate": 0.0009864301666940183, "loss": 1.4854, "step": 5267 }, { "epoch": 0.52, "learning_rate": 0.0009861108967952634, "loss": 1.46, "step": 5268 }, { "epoch": 0.52, "learning_rate": 0.0009857916283125374, "loss": 1.4473, "step": 5269 }, { "epoch": 0.52, "learning_rate": 0.00098547236127839, "loss": 1.5049, "step": 5270 }, { "epoch": 0.52, "learning_rate": 0.0009851530957253715, "loss": 1.457, "step": 5271 }, { "epoch": 0.52, "learning_rate": 0.0009848338316860317, "loss": 1.5078, "step": 5272 }, { "epoch": 0.52, "learning_rate": 0.0009845145691929206, "loss": 1.5811, "step": 5273 }, { "epoch": 0.52, "learning_rate": 0.0009841953082785875, "loss": 1.6689, "step": 5274 }, { "epoch": 0.52, "learning_rate": 0.0009838760489755823, "loss": 1.502, "step": 5275 }, { "epoch": 0.52, "learning_rate": 0.0009835567913164538, "loss": 1.5156, "step": 5276 }, { "epoch": 0.52, "learning_rate": 0.0009832375353337514, "loss": 1.5439, "step": 5277 }, { "epoch": 0.52, "learning_rate": 0.000982918281060024, "loss": 1.6377, "step": 5278 }, { "epoch": 0.52, "learning_rate": 0.00098259902852782, "loss": 1.6006, "step": 5279 }, { "epoch": 0.52, "learning_rate": 0.0009822797777696887, "loss": 1.5137, "step": 5280 }, { "epoch": 0.52, "learning_rate": 0.0009819605288181783, "loss": 1.5879, "step": 5281 }, { "epoch": 0.52, "learning_rate": 0.0009816412817058363, "loss": 1.4326, "step": 5282 }, { "epoch": 0.52, "learning_rate": 0.0009813220364652115, "loss": 1.5781, "step": 5283 }, { "epoch": 0.52, "learning_rate": 0.0009810027931288516, "loss": 1.4834, "step": 5284 }, { "epoch": 0.52, "learning_rate": 0.000980683551729304, "loss": 1.457, "step": 5285 }, { "epoch": 0.52, "learning_rate": 0.0009803643122991163, "loss": 1.4443, "step": 5286 }, { "epoch": 0.52, "learning_rate": 0.0009800450748708356, "loss": 1.5352, "step": 5287 }, { "epoch": 0.52, "learning_rate": 0.0009797258394770095, "loss": 1.4863, "step": 5288 }, { "epoch": 0.52, "learning_rate": 0.0009794066061501842, "loss": 1.6338, "step": 5289 }, { "epoch": 0.52, "learning_rate": 0.0009790873749229068, "loss": 1.4131, "step": 5290 }, { "epoch": 0.52, "learning_rate": 0.0009787681458277235, "loss": 1.6133, "step": 5291 }, { "epoch": 0.52, "learning_rate": 0.0009784489188971804, "loss": 1.5645, "step": 5292 }, { "epoch": 0.52, "learning_rate": 0.0009781296941638238, "loss": 1.4922, "step": 5293 }, { "epoch": 0.52, "learning_rate": 0.0009778104716601991, "loss": 1.4912, "step": 5294 }, { "epoch": 0.52, "learning_rate": 0.000977491251418852, "loss": 1.623, "step": 5295 }, { "epoch": 0.52, "learning_rate": 0.0009771720334723275, "loss": 1.5244, "step": 5296 }, { "epoch": 0.52, "learning_rate": 0.0009768528178531713, "loss": 1.4717, "step": 5297 }, { "epoch": 0.52, "learning_rate": 0.0009765336045939277, "loss": 1.5381, "step": 5298 }, { "epoch": 0.52, "learning_rate": 0.0009762143937271416, "loss": 1.4404, "step": 5299 }, { "epoch": 0.52, "learning_rate": 0.0009758951852853571, "loss": 1.5645, "step": 5300 }, { "epoch": 0.52, "learning_rate": 0.0009755759793011185, "loss": 1.6699, "step": 5301 }, { "epoch": 0.52, "learning_rate": 0.0009752567758069694, "loss": 1.5615, "step": 5302 }, { "epoch": 0.52, "learning_rate": 0.0009749375748354536, "loss": 1.4639, "step": 5303 }, { "epoch": 0.52, "learning_rate": 0.0009746183764191143, "loss": 1.5342, "step": 5304 }, { "epoch": 0.52, "learning_rate": 0.0009742991805904945, "loss": 1.542, "step": 5305 }, { "epoch": 0.52, "learning_rate": 0.0009739799873821373, "loss": 1.4004, "step": 5306 }, { "epoch": 0.52, "learning_rate": 0.0009736607968265851, "loss": 1.6611, "step": 5307 }, { "epoch": 0.52, "learning_rate": 0.00097334160895638, "loss": 1.6006, "step": 5308 }, { "epoch": 0.52, "learning_rate": 0.0009730224238040644, "loss": 1.4341, "step": 5309 }, { "epoch": 0.52, "learning_rate": 0.0009727032414021796, "loss": 1.5625, "step": 5310 }, { "epoch": 0.52, "learning_rate": 0.0009723840617832672, "loss": 1.5137, "step": 5311 }, { "epoch": 0.52, "learning_rate": 0.0009720648849798684, "loss": 1.5234, "step": 5312 }, { "epoch": 0.52, "learning_rate": 0.000971745711024524, "loss": 1.498, "step": 5313 }, { "epoch": 0.52, "learning_rate": 0.0009714265399497744, "loss": 1.4219, "step": 5314 }, { "epoch": 0.52, "learning_rate": 0.0009711073717881602, "loss": 1.4912, "step": 5315 }, { "epoch": 0.52, "learning_rate": 0.0009707882065722214, "loss": 1.5342, "step": 5316 }, { "epoch": 0.52, "learning_rate": 0.0009704690443344976, "loss": 1.5977, "step": 5317 }, { "epoch": 0.52, "learning_rate": 0.0009701498851075281, "loss": 1.5703, "step": 5318 }, { "epoch": 0.52, "learning_rate": 0.0009698307289238519, "loss": 1.6094, "step": 5319 }, { "epoch": 0.52, "learning_rate": 0.0009695115758160081, "loss": 1.6602, "step": 5320 }, { "epoch": 0.52, "learning_rate": 0.0009691924258165348, "loss": 1.4834, "step": 5321 }, { "epoch": 0.52, "learning_rate": 0.0009688732789579704, "loss": 1.6387, "step": 5322 }, { "epoch": 0.52, "learning_rate": 0.0009685541352728523, "loss": 1.5117, "step": 5323 }, { "epoch": 0.52, "learning_rate": 0.0009682349947937184, "loss": 1.4883, "step": 5324 }, { "epoch": 0.52, "learning_rate": 0.000967915857553106, "loss": 1.5791, "step": 5325 }, { "epoch": 0.53, "learning_rate": 0.0009675967235835514, "loss": 1.5977, "step": 5326 }, { "epoch": 0.53, "learning_rate": 0.0009672775929175917, "loss": 1.5312, "step": 5327 }, { "epoch": 0.53, "learning_rate": 0.0009669584655877625, "loss": 1.5215, "step": 5328 }, { "epoch": 0.53, "learning_rate": 0.0009666393416266, "loss": 1.5293, "step": 5329 }, { "epoch": 0.53, "learning_rate": 0.0009663202210666395, "loss": 1.4014, "step": 5330 }, { "epoch": 0.53, "learning_rate": 0.0009660011039404162, "loss": 1.4092, "step": 5331 }, { "epoch": 0.53, "learning_rate": 0.0009656819902804648, "loss": 1.4932, "step": 5332 }, { "epoch": 0.53, "learning_rate": 0.0009653628801193199, "loss": 1.5693, "step": 5333 }, { "epoch": 0.53, "learning_rate": 0.0009650437734895154, "loss": 1.541, "step": 5334 }, { "epoch": 0.53, "learning_rate": 0.0009647246704235853, "loss": 1.4658, "step": 5335 }, { "epoch": 0.53, "learning_rate": 0.0009644055709540626, "loss": 1.4727, "step": 5336 }, { "epoch": 0.53, "learning_rate": 0.0009640864751134806, "loss": 1.4951, "step": 5337 }, { "epoch": 0.53, "learning_rate": 0.0009637673829343717, "loss": 1.5137, "step": 5338 }, { "epoch": 0.53, "learning_rate": 0.000963448294449268, "loss": 1.5098, "step": 5339 }, { "epoch": 0.53, "learning_rate": 0.0009631292096907018, "loss": 1.5889, "step": 5340 }, { "epoch": 0.53, "learning_rate": 0.0009628101286912042, "loss": 1.6064, "step": 5341 }, { "epoch": 0.53, "learning_rate": 0.0009624910514833064, "loss": 1.5, "step": 5342 }, { "epoch": 0.53, "learning_rate": 0.0009621719780995392, "loss": 1.5381, "step": 5343 }, { "epoch": 0.53, "learning_rate": 0.0009618529085724329, "loss": 1.5381, "step": 5344 }, { "epoch": 0.53, "learning_rate": 0.0009615338429345174, "loss": 1.2822, "step": 5345 }, { "epoch": 0.53, "learning_rate": 0.000961214781218322, "loss": 1.5898, "step": 5346 }, { "epoch": 0.53, "learning_rate": 0.0009608957234563763, "loss": 1.5049, "step": 5347 }, { "epoch": 0.53, "learning_rate": 0.0009605766696812088, "loss": 1.5352, "step": 5348 }, { "epoch": 0.53, "learning_rate": 0.0009602576199253479, "loss": 1.4521, "step": 5349 }, { "epoch": 0.53, "learning_rate": 0.0009599385742213213, "loss": 1.5762, "step": 5350 }, { "epoch": 0.53, "learning_rate": 0.0009596195326016568, "loss": 1.5137, "step": 5351 }, { "epoch": 0.53, "learning_rate": 0.0009593004950988814, "loss": 1.6846, "step": 5352 }, { "epoch": 0.53, "learning_rate": 0.0009589814617455216, "loss": 1.4937, "step": 5353 }, { "epoch": 0.53, "learning_rate": 0.0009586624325741038, "loss": 1.6328, "step": 5354 }, { "epoch": 0.53, "learning_rate": 0.0009583434076171534, "loss": 1.4668, "step": 5355 }, { "epoch": 0.53, "learning_rate": 0.0009580243869071965, "loss": 1.6143, "step": 5356 }, { "epoch": 0.53, "learning_rate": 0.0009577053704767578, "loss": 1.3896, "step": 5357 }, { "epoch": 0.53, "learning_rate": 0.0009573863583583616, "loss": 1.5107, "step": 5358 }, { "epoch": 0.53, "learning_rate": 0.0009570673505845321, "loss": 1.5547, "step": 5359 }, { "epoch": 0.53, "learning_rate": 0.000956748347187793, "loss": 1.5088, "step": 5360 }, { "epoch": 0.53, "learning_rate": 0.0009564293482006674, "loss": 1.4619, "step": 5361 }, { "epoch": 0.53, "learning_rate": 0.0009561103536556781, "loss": 1.5723, "step": 5362 }, { "epoch": 0.53, "learning_rate": 0.0009557913635853474, "loss": 1.5078, "step": 5363 }, { "epoch": 0.53, "learning_rate": 0.0009554723780221967, "loss": 1.5322, "step": 5364 }, { "epoch": 0.53, "learning_rate": 0.0009551533969987483, "loss": 1.5107, "step": 5365 }, { "epoch": 0.53, "learning_rate": 0.0009548344205475224, "loss": 1.3745, "step": 5366 }, { "epoch": 0.53, "learning_rate": 0.0009545154487010395, "loss": 1.4883, "step": 5367 }, { "epoch": 0.53, "learning_rate": 0.0009541964814918197, "loss": 1.6279, "step": 5368 }, { "epoch": 0.53, "learning_rate": 0.0009538775189523826, "loss": 1.5732, "step": 5369 }, { "epoch": 0.53, "learning_rate": 0.0009535585611152471, "loss": 1.543, "step": 5370 }, { "epoch": 0.53, "learning_rate": 0.0009532396080129316, "loss": 1.5898, "step": 5371 }, { "epoch": 0.53, "learning_rate": 0.0009529206596779545, "loss": 1.4492, "step": 5372 }, { "epoch": 0.53, "learning_rate": 0.0009526017161428328, "loss": 1.5059, "step": 5373 }, { "epoch": 0.53, "learning_rate": 0.0009522827774400844, "loss": 1.4497, "step": 5374 }, { "epoch": 0.53, "learning_rate": 0.0009519638436022254, "loss": 1.5376, "step": 5375 }, { "epoch": 0.53, "learning_rate": 0.000951644914661772, "loss": 1.6133, "step": 5376 }, { "epoch": 0.53, "learning_rate": 0.0009513259906512397, "loss": 1.5254, "step": 5377 }, { "epoch": 0.53, "learning_rate": 0.0009510070716031436, "loss": 1.6211, "step": 5378 }, { "epoch": 0.53, "learning_rate": 0.0009506881575499985, "loss": 1.6748, "step": 5379 }, { "epoch": 0.53, "learning_rate": 0.0009503692485243182, "loss": 1.5332, "step": 5380 }, { "epoch": 0.53, "learning_rate": 0.0009500503445586165, "loss": 1.5801, "step": 5381 }, { "epoch": 0.53, "learning_rate": 0.000949731445685406, "loss": 1.6885, "step": 5382 }, { "epoch": 0.53, "learning_rate": 0.0009494125519371999, "loss": 1.6279, "step": 5383 }, { "epoch": 0.53, "learning_rate": 0.0009490936633465099, "loss": 1.5771, "step": 5384 }, { "epoch": 0.53, "learning_rate": 0.0009487747799458475, "loss": 1.376, "step": 5385 }, { "epoch": 0.53, "learning_rate": 0.0009484559017677237, "loss": 1.4697, "step": 5386 }, { "epoch": 0.53, "learning_rate": 0.0009481370288446487, "loss": 1.6387, "step": 5387 }, { "epoch": 0.53, "learning_rate": 0.0009478181612091325, "loss": 1.4678, "step": 5388 }, { "epoch": 0.53, "learning_rate": 0.0009474992988936847, "loss": 1.5869, "step": 5389 }, { "epoch": 0.53, "learning_rate": 0.0009471804419308138, "loss": 1.5059, "step": 5390 }, { "epoch": 0.53, "learning_rate": 0.000946861590353028, "loss": 1.5703, "step": 5391 }, { "epoch": 0.53, "learning_rate": 0.0009465427441928354, "loss": 1.4424, "step": 5392 }, { "epoch": 0.53, "learning_rate": 0.0009462239034827428, "loss": 1.4609, "step": 5393 }, { "epoch": 0.53, "learning_rate": 0.0009459050682552575, "loss": 1.54, "step": 5394 }, { "epoch": 0.53, "learning_rate": 0.0009455862385428846, "loss": 1.541, "step": 5395 }, { "epoch": 0.53, "learning_rate": 0.0009452674143781294, "loss": 1.4434, "step": 5396 }, { "epoch": 0.53, "learning_rate": 0.0009449485957934979, "loss": 1.4404, "step": 5397 }, { "epoch": 0.53, "learning_rate": 0.0009446297828214938, "loss": 1.6025, "step": 5398 }, { "epoch": 0.53, "learning_rate": 0.0009443109754946211, "loss": 1.4512, "step": 5399 }, { "epoch": 0.53, "learning_rate": 0.0009439921738453828, "loss": 1.4102, "step": 5400 }, { "epoch": 0.53, "learning_rate": 0.0009436733779062817, "loss": 1.5322, "step": 5401 }, { "epoch": 0.53, "learning_rate": 0.0009433545877098197, "loss": 1.4326, "step": 5402 }, { "epoch": 0.53, "learning_rate": 0.0009430358032884984, "loss": 1.4941, "step": 5403 }, { "epoch": 0.53, "learning_rate": 0.0009427170246748184, "loss": 1.5918, "step": 5404 }, { "epoch": 0.53, "learning_rate": 0.0009423982519012801, "loss": 1.5996, "step": 5405 }, { "epoch": 0.53, "learning_rate": 0.0009420794850003834, "loss": 1.6006, "step": 5406 }, { "epoch": 0.53, "learning_rate": 0.0009417607240046272, "loss": 1.6982, "step": 5407 }, { "epoch": 0.53, "learning_rate": 0.0009414419689465099, "loss": 1.541, "step": 5408 }, { "epoch": 0.53, "learning_rate": 0.0009411232198585296, "loss": 1.5771, "step": 5409 }, { "epoch": 0.53, "learning_rate": 0.0009408044767731833, "loss": 1.5459, "step": 5410 }, { "epoch": 0.53, "learning_rate": 0.0009404857397229677, "loss": 1.6318, "step": 5411 }, { "epoch": 0.53, "learning_rate": 0.000940167008740379, "loss": 1.4355, "step": 5412 }, { "epoch": 0.53, "learning_rate": 0.0009398482838579124, "loss": 1.5342, "step": 5413 }, { "epoch": 0.53, "learning_rate": 0.0009395295651080627, "loss": 1.5938, "step": 5414 }, { "epoch": 0.53, "learning_rate": 0.0009392108525233244, "loss": 1.3984, "step": 5415 }, { "epoch": 0.53, "learning_rate": 0.0009388921461361908, "loss": 1.4873, "step": 5416 }, { "epoch": 0.53, "learning_rate": 0.0009385734459791547, "loss": 1.6035, "step": 5417 }, { "epoch": 0.53, "learning_rate": 0.0009382547520847087, "loss": 1.6045, "step": 5418 }, { "epoch": 0.53, "learning_rate": 0.000937936064485344, "loss": 1.3984, "step": 5419 }, { "epoch": 0.53, "learning_rate": 0.0009376173832135521, "loss": 1.4834, "step": 5420 }, { "epoch": 0.53, "learning_rate": 0.0009372987083018228, "loss": 1.5605, "step": 5421 }, { "epoch": 0.53, "learning_rate": 0.0009369800397826463, "loss": 1.4639, "step": 5422 }, { "epoch": 0.53, "learning_rate": 0.000936661377688511, "loss": 1.5264, "step": 5423 }, { "epoch": 0.53, "learning_rate": 0.0009363427220519061, "loss": 1.5781, "step": 5424 }, { "epoch": 0.53, "learning_rate": 0.0009360240729053188, "loss": 1.46, "step": 5425 }, { "epoch": 0.53, "learning_rate": 0.0009357054302812364, "loss": 1.4404, "step": 5426 }, { "epoch": 0.53, "learning_rate": 0.0009353867942121452, "loss": 1.6055, "step": 5427 }, { "epoch": 0.54, "learning_rate": 0.000935068164730531, "loss": 1.6055, "step": 5428 }, { "epoch": 0.54, "learning_rate": 0.0009347495418688789, "loss": 1.3906, "step": 5429 }, { "epoch": 0.54, "learning_rate": 0.000934430925659673, "loss": 1.6416, "step": 5430 }, { "epoch": 0.54, "learning_rate": 0.0009341123161353974, "loss": 1.4238, "step": 5431 }, { "epoch": 0.54, "learning_rate": 0.0009337937133285348, "loss": 1.4717, "step": 5432 }, { "epoch": 0.54, "learning_rate": 0.0009334751172715679, "loss": 1.3984, "step": 5433 }, { "epoch": 0.54, "learning_rate": 0.0009331565279969782, "loss": 1.5879, "step": 5434 }, { "epoch": 0.54, "learning_rate": 0.0009328379455372465, "loss": 1.4912, "step": 5435 }, { "epoch": 0.54, "learning_rate": 0.0009325193699248534, "loss": 1.6221, "step": 5436 }, { "epoch": 0.54, "learning_rate": 0.0009322008011922782, "loss": 1.4834, "step": 5437 }, { "epoch": 0.54, "learning_rate": 0.000931882239372, "loss": 1.5, "step": 5438 }, { "epoch": 0.54, "learning_rate": 0.0009315636844964968, "loss": 1.5186, "step": 5439 }, { "epoch": 0.54, "learning_rate": 0.000931245136598246, "loss": 1.5811, "step": 5440 }, { "epoch": 0.54, "learning_rate": 0.0009309265957097243, "loss": 1.5684, "step": 5441 }, { "epoch": 0.54, "learning_rate": 0.000930608061863408, "loss": 1.458, "step": 5442 }, { "epoch": 0.54, "learning_rate": 0.0009302895350917726, "loss": 1.4424, "step": 5443 }, { "epoch": 0.54, "learning_rate": 0.0009299710154272922, "loss": 1.5986, "step": 5444 }, { "epoch": 0.54, "learning_rate": 0.0009296525029024409, "loss": 1.4229, "step": 5445 }, { "epoch": 0.54, "learning_rate": 0.0009293339975496918, "loss": 1.5312, "step": 5446 }, { "epoch": 0.54, "learning_rate": 0.0009290154994015174, "loss": 1.5566, "step": 5447 }, { "epoch": 0.54, "learning_rate": 0.0009286970084903893, "loss": 1.4199, "step": 5448 }, { "epoch": 0.54, "learning_rate": 0.0009283785248487784, "loss": 1.416, "step": 5449 }, { "epoch": 0.54, "learning_rate": 0.0009280600485091552, "loss": 1.3892, "step": 5450 }, { "epoch": 0.54, "learning_rate": 0.0009277415795039888, "loss": 1.5859, "step": 5451 }, { "epoch": 0.54, "learning_rate": 0.0009274231178657479, "loss": 1.5449, "step": 5452 }, { "epoch": 0.54, "learning_rate": 0.0009271046636269005, "loss": 1.5215, "step": 5453 }, { "epoch": 0.54, "learning_rate": 0.0009267862168199139, "loss": 1.6992, "step": 5454 }, { "epoch": 0.54, "learning_rate": 0.0009264677774772546, "loss": 1.4326, "step": 5455 }, { "epoch": 0.54, "learning_rate": 0.000926149345631388, "loss": 1.4434, "step": 5456 }, { "epoch": 0.54, "learning_rate": 0.0009258309213147793, "loss": 1.5508, "step": 5457 }, { "epoch": 0.54, "learning_rate": 0.0009255125045598926, "loss": 1.5244, "step": 5458 }, { "epoch": 0.54, "learning_rate": 0.0009251940953991913, "loss": 1.4307, "step": 5459 }, { "epoch": 0.54, "learning_rate": 0.0009248756938651379, "loss": 1.6396, "step": 5460 }, { "epoch": 0.54, "learning_rate": 0.0009245572999901944, "loss": 1.4531, "step": 5461 }, { "epoch": 0.54, "learning_rate": 0.0009242389138068215, "loss": 1.5371, "step": 5462 }, { "epoch": 0.54, "learning_rate": 0.0009239205353474798, "loss": 1.5361, "step": 5463 }, { "epoch": 0.54, "learning_rate": 0.0009236021646446288, "loss": 1.5225, "step": 5464 }, { "epoch": 0.54, "learning_rate": 0.0009232838017307267, "loss": 1.3721, "step": 5465 }, { "epoch": 0.54, "learning_rate": 0.000922965446638232, "loss": 1.4893, "step": 5466 }, { "epoch": 0.54, "learning_rate": 0.0009226470993996015, "loss": 1.4492, "step": 5467 }, { "epoch": 0.54, "learning_rate": 0.0009223287600472915, "loss": 1.3428, "step": 5468 }, { "epoch": 0.54, "learning_rate": 0.0009220104286137576, "loss": 1.4736, "step": 5469 }, { "epoch": 0.54, "learning_rate": 0.0009216921051314543, "loss": 1.4893, "step": 5470 }, { "epoch": 0.54, "learning_rate": 0.0009213737896328355, "loss": 1.5049, "step": 5471 }, { "epoch": 0.54, "learning_rate": 0.0009210554821503545, "loss": 1.5, "step": 5472 }, { "epoch": 0.54, "learning_rate": 0.0009207371827164633, "loss": 1.5635, "step": 5473 }, { "epoch": 0.54, "learning_rate": 0.0009204188913636131, "loss": 1.5283, "step": 5474 }, { "epoch": 0.54, "learning_rate": 0.0009201006081242551, "loss": 1.3945, "step": 5475 }, { "epoch": 0.54, "learning_rate": 0.0009197823330308386, "loss": 1.5288, "step": 5476 }, { "epoch": 0.54, "learning_rate": 0.000919464066115813, "loss": 1.5576, "step": 5477 }, { "epoch": 0.54, "learning_rate": 0.0009191458074116256, "loss": 1.6455, "step": 5478 }, { "epoch": 0.54, "learning_rate": 0.0009188275569507245, "loss": 1.4922, "step": 5479 }, { "epoch": 0.54, "learning_rate": 0.0009185093147655557, "loss": 1.4863, "step": 5480 }, { "epoch": 0.54, "learning_rate": 0.0009181910808885648, "loss": 1.5254, "step": 5481 }, { "epoch": 0.54, "learning_rate": 0.0009178728553521965, "loss": 1.5654, "step": 5482 }, { "epoch": 0.54, "learning_rate": 0.0009175546381888947, "loss": 1.4404, "step": 5483 }, { "epoch": 0.54, "learning_rate": 0.0009172364294311027, "loss": 1.5293, "step": 5484 }, { "epoch": 0.54, "learning_rate": 0.0009169182291112626, "loss": 1.6562, "step": 5485 }, { "epoch": 0.54, "learning_rate": 0.0009166000372618153, "loss": 1.5645, "step": 5486 }, { "epoch": 0.54, "learning_rate": 0.0009162818539152018, "loss": 1.4092, "step": 5487 }, { "epoch": 0.54, "learning_rate": 0.0009159636791038613, "loss": 1.6719, "step": 5488 }, { "epoch": 0.54, "learning_rate": 0.0009156455128602327, "loss": 1.5117, "step": 5489 }, { "epoch": 0.54, "learning_rate": 0.0009153273552167536, "loss": 1.4941, "step": 5490 }, { "epoch": 0.54, "learning_rate": 0.0009150092062058614, "loss": 1.4834, "step": 5491 }, { "epoch": 0.54, "learning_rate": 0.0009146910658599916, "loss": 1.5645, "step": 5492 }, { "epoch": 0.54, "learning_rate": 0.0009143729342115799, "loss": 1.5088, "step": 5493 }, { "epoch": 0.54, "learning_rate": 0.0009140548112930606, "loss": 1.6094, "step": 5494 }, { "epoch": 0.54, "learning_rate": 0.0009137366971368669, "loss": 1.5225, "step": 5495 }, { "epoch": 0.54, "learning_rate": 0.0009134185917754312, "loss": 1.5859, "step": 5496 }, { "epoch": 0.54, "learning_rate": 0.0009131004952411854, "loss": 1.4395, "step": 5497 }, { "epoch": 0.54, "learning_rate": 0.0009127824075665603, "loss": 1.6172, "step": 5498 }, { "epoch": 0.54, "learning_rate": 0.0009124643287839855, "loss": 1.5293, "step": 5499 }, { "epoch": 0.54, "learning_rate": 0.0009121462589258899, "loss": 1.6045, "step": 5500 }, { "epoch": 0.54, "learning_rate": 0.0009118281980247014, "loss": 1.6191, "step": 5501 }, { "epoch": 0.54, "learning_rate": 0.0009115101461128477, "loss": 1.4688, "step": 5502 }, { "epoch": 0.54, "learning_rate": 0.0009111921032227545, "loss": 1.3809, "step": 5503 }, { "epoch": 0.54, "learning_rate": 0.0009108740693868472, "loss": 1.582, "step": 5504 }, { "epoch": 0.54, "learning_rate": 0.00091055604463755, "loss": 1.5615, "step": 5505 }, { "epoch": 0.54, "learning_rate": 0.0009102380290072868, "loss": 1.5596, "step": 5506 }, { "epoch": 0.54, "learning_rate": 0.0009099200225284793, "loss": 1.5371, "step": 5507 }, { "epoch": 0.54, "learning_rate": 0.0009096020252335496, "loss": 1.5303, "step": 5508 }, { "epoch": 0.54, "learning_rate": 0.0009092840371549183, "loss": 1.417, "step": 5509 }, { "epoch": 0.54, "learning_rate": 0.0009089660583250046, "loss": 1.5801, "step": 5510 }, { "epoch": 0.54, "learning_rate": 0.0009086480887762279, "loss": 1.3906, "step": 5511 }, { "epoch": 0.54, "learning_rate": 0.0009083301285410054, "loss": 1.5391, "step": 5512 }, { "epoch": 0.54, "learning_rate": 0.0009080121776517543, "loss": 1.5732, "step": 5513 }, { "epoch": 0.54, "learning_rate": 0.0009076942361408904, "loss": 1.6035, "step": 5514 }, { "epoch": 0.54, "learning_rate": 0.0009073763040408284, "loss": 1.5625, "step": 5515 }, { "epoch": 0.54, "learning_rate": 0.0009070583813839826, "loss": 1.4678, "step": 5516 }, { "epoch": 0.54, "learning_rate": 0.000906740468202766, "loss": 1.4727, "step": 5517 }, { "epoch": 0.54, "learning_rate": 0.0009064225645295904, "loss": 1.4688, "step": 5518 }, { "epoch": 0.54, "learning_rate": 0.0009061046703968672, "loss": 1.4238, "step": 5519 }, { "epoch": 0.54, "learning_rate": 0.0009057867858370061, "loss": 1.2754, "step": 5520 }, { "epoch": 0.54, "learning_rate": 0.0009054689108824165, "loss": 1.541, "step": 5521 }, { "epoch": 0.54, "learning_rate": 0.0009051510455655064, "loss": 1.5068, "step": 5522 }, { "epoch": 0.54, "learning_rate": 0.0009048331899186831, "loss": 1.5078, "step": 5523 }, { "epoch": 0.54, "learning_rate": 0.0009045153439743523, "loss": 1.5137, "step": 5524 }, { "epoch": 0.54, "learning_rate": 0.0009041975077649199, "loss": 1.5459, "step": 5525 }, { "epoch": 0.54, "learning_rate": 0.0009038796813227898, "loss": 1.5615, "step": 5526 }, { "epoch": 0.54, "learning_rate": 0.0009035618646803651, "loss": 1.5303, "step": 5527 }, { "epoch": 0.54, "learning_rate": 0.0009032440578700479, "loss": 1.5107, "step": 5528 }, { "epoch": 0.55, "learning_rate": 0.0009029262609242398, "loss": 1.4932, "step": 5529 }, { "epoch": 0.55, "learning_rate": 0.0009026084738753406, "loss": 1.4775, "step": 5530 }, { "epoch": 0.55, "learning_rate": 0.0009022906967557496, "loss": 1.5176, "step": 5531 }, { "epoch": 0.55, "learning_rate": 0.0009019729295978651, "loss": 1.5225, "step": 5532 }, { "epoch": 0.55, "learning_rate": 0.0009016551724340838, "loss": 1.5186, "step": 5533 }, { "epoch": 0.55, "learning_rate": 0.0009013374252968024, "loss": 1.5166, "step": 5534 }, { "epoch": 0.55, "learning_rate": 0.0009010196882184158, "loss": 1.459, "step": 5535 }, { "epoch": 0.55, "learning_rate": 0.0009007019612313181, "loss": 1.5811, "step": 5536 }, { "epoch": 0.55, "learning_rate": 0.0009003842443679024, "loss": 1.5859, "step": 5537 }, { "epoch": 0.55, "learning_rate": 0.0009000665376605604, "loss": 1.4971, "step": 5538 }, { "epoch": 0.55, "learning_rate": 0.0008997488411416836, "loss": 1.4824, "step": 5539 }, { "epoch": 0.55, "learning_rate": 0.0008994311548436615, "loss": 1.5273, "step": 5540 }, { "epoch": 0.55, "learning_rate": 0.0008991134787988831, "loss": 1.5566, "step": 5541 }, { "epoch": 0.55, "learning_rate": 0.0008987958130397363, "loss": 1.5361, "step": 5542 }, { "epoch": 0.55, "learning_rate": 0.0008984781575986083, "loss": 1.6025, "step": 5543 }, { "epoch": 0.55, "learning_rate": 0.0008981605125078843, "loss": 1.5146, "step": 5544 }, { "epoch": 0.55, "learning_rate": 0.0008978428777999493, "loss": 1.4551, "step": 5545 }, { "epoch": 0.55, "learning_rate": 0.000897525253507187, "loss": 1.4756, "step": 5546 }, { "epoch": 0.55, "learning_rate": 0.0008972076396619798, "loss": 1.4893, "step": 5547 }, { "epoch": 0.55, "learning_rate": 0.0008968900362967091, "loss": 1.5039, "step": 5548 }, { "epoch": 0.55, "learning_rate": 0.0008965724434437558, "loss": 1.6104, "step": 5549 }, { "epoch": 0.55, "learning_rate": 0.0008962548611354989, "loss": 1.4932, "step": 5550 }, { "epoch": 0.55, "learning_rate": 0.0008959372894043168, "loss": 1.5732, "step": 5551 }, { "epoch": 0.55, "learning_rate": 0.0008956197282825866, "loss": 1.4736, "step": 5552 }, { "epoch": 0.55, "learning_rate": 0.000895302177802685, "loss": 1.4883, "step": 5553 }, { "epoch": 0.55, "learning_rate": 0.0008949846379969863, "loss": 1.543, "step": 5554 }, { "epoch": 0.55, "learning_rate": 0.0008946671088978651, "loss": 1.4785, "step": 5555 }, { "epoch": 0.55, "learning_rate": 0.000894349590537694, "loss": 1.4824, "step": 5556 }, { "epoch": 0.55, "learning_rate": 0.0008940320829488447, "loss": 1.4834, "step": 5557 }, { "epoch": 0.55, "learning_rate": 0.0008937145861636881, "loss": 1.4609, "step": 5558 }, { "epoch": 0.55, "learning_rate": 0.0008933971002145937, "loss": 1.5664, "step": 5559 }, { "epoch": 0.55, "learning_rate": 0.0008930796251339296, "loss": 1.5566, "step": 5560 }, { "epoch": 0.55, "learning_rate": 0.0008927621609540638, "loss": 1.5625, "step": 5561 }, { "epoch": 0.55, "learning_rate": 0.0008924447077073627, "loss": 1.4551, "step": 5562 }, { "epoch": 0.55, "learning_rate": 0.0008921272654261907, "loss": 1.5449, "step": 5563 }, { "epoch": 0.55, "learning_rate": 0.0008918098341429122, "loss": 1.6064, "step": 5564 }, { "epoch": 0.55, "learning_rate": 0.0008914924138898897, "loss": 1.54, "step": 5565 }, { "epoch": 0.55, "learning_rate": 0.0008911750046994857, "loss": 1.5, "step": 5566 }, { "epoch": 0.55, "learning_rate": 0.0008908576066040606, "loss": 1.4385, "step": 5567 }, { "epoch": 0.55, "learning_rate": 0.0008905402196359737, "loss": 1.4541, "step": 5568 }, { "epoch": 0.55, "learning_rate": 0.0008902228438275835, "loss": 1.3975, "step": 5569 }, { "epoch": 0.55, "learning_rate": 0.0008899054792112475, "loss": 1.5508, "step": 5570 }, { "epoch": 0.55, "learning_rate": 0.0008895881258193215, "loss": 1.4131, "step": 5571 }, { "epoch": 0.55, "learning_rate": 0.0008892707836841605, "loss": 1.6396, "step": 5572 }, { "epoch": 0.55, "learning_rate": 0.0008889534528381185, "loss": 1.5283, "step": 5573 }, { "epoch": 0.55, "learning_rate": 0.0008886361333135477, "loss": 1.4521, "step": 5574 }, { "epoch": 0.55, "learning_rate": 0.0008883188251428003, "loss": 1.4463, "step": 5575 }, { "epoch": 0.55, "learning_rate": 0.0008880015283582263, "loss": 1.4932, "step": 5576 }, { "epoch": 0.55, "learning_rate": 0.000887684242992175, "loss": 1.5898, "step": 5577 }, { "epoch": 0.55, "learning_rate": 0.0008873669690769942, "loss": 1.5928, "step": 5578 }, { "epoch": 0.55, "learning_rate": 0.000887049706645031, "loss": 1.3691, "step": 5579 }, { "epoch": 0.55, "learning_rate": 0.000886732455728631, "loss": 1.5225, "step": 5580 }, { "epoch": 0.55, "learning_rate": 0.0008864152163601389, "loss": 1.4678, "step": 5581 }, { "epoch": 0.55, "learning_rate": 0.0008860979885718977, "loss": 1.4365, "step": 5582 }, { "epoch": 0.55, "learning_rate": 0.0008857807723962496, "loss": 1.4346, "step": 5583 }, { "epoch": 0.55, "learning_rate": 0.0008854635678655358, "loss": 1.6006, "step": 5584 }, { "epoch": 0.55, "learning_rate": 0.0008851463750120962, "loss": 1.5225, "step": 5585 }, { "epoch": 0.55, "learning_rate": 0.000884829193868269, "loss": 1.5342, "step": 5586 }, { "epoch": 0.55, "learning_rate": 0.000884512024466392, "loss": 1.6172, "step": 5587 }, { "epoch": 0.55, "learning_rate": 0.0008841948668388011, "loss": 1.4893, "step": 5588 }, { "epoch": 0.55, "learning_rate": 0.0008838777210178315, "loss": 1.5303, "step": 5589 }, { "epoch": 0.55, "learning_rate": 0.000883560587035817, "loss": 1.4453, "step": 5590 }, { "epoch": 0.55, "learning_rate": 0.00088324346492509, "loss": 1.584, "step": 5591 }, { "epoch": 0.55, "learning_rate": 0.0008829263547179819, "loss": 1.5098, "step": 5592 }, { "epoch": 0.55, "learning_rate": 0.0008826092564468231, "loss": 1.4766, "step": 5593 }, { "epoch": 0.55, "learning_rate": 0.0008822921701439426, "loss": 1.5098, "step": 5594 }, { "epoch": 0.55, "learning_rate": 0.000881975095841668, "loss": 1.4414, "step": 5595 }, { "epoch": 0.55, "learning_rate": 0.0008816580335723256, "loss": 1.4854, "step": 5596 }, { "epoch": 0.55, "learning_rate": 0.000881340983368241, "loss": 1.543, "step": 5597 }, { "epoch": 0.55, "learning_rate": 0.0008810239452617381, "loss": 1.4902, "step": 5598 }, { "epoch": 0.55, "learning_rate": 0.0008807069192851397, "loss": 1.5322, "step": 5599 }, { "epoch": 0.55, "learning_rate": 0.0008803899054707677, "loss": 1.502, "step": 5600 }, { "epoch": 0.55, "learning_rate": 0.0008800729038509419, "loss": 1.5381, "step": 5601 }, { "epoch": 0.55, "learning_rate": 0.0008797559144579819, "loss": 1.5, "step": 5602 }, { "epoch": 0.55, "learning_rate": 0.0008794389373242053, "loss": 1.4209, "step": 5603 }, { "epoch": 0.55, "learning_rate": 0.0008791219724819288, "loss": 1.6289, "step": 5604 }, { "epoch": 0.55, "learning_rate": 0.0008788050199634677, "loss": 1.5752, "step": 5605 }, { "epoch": 0.55, "learning_rate": 0.0008784880798011363, "loss": 1.5869, "step": 5606 }, { "epoch": 0.55, "learning_rate": 0.0008781711520272471, "loss": 1.5811, "step": 5607 }, { "epoch": 0.55, "learning_rate": 0.0008778542366741118, "loss": 1.5312, "step": 5608 }, { "epoch": 0.55, "learning_rate": 0.0008775373337740408, "loss": 1.5459, "step": 5609 }, { "epoch": 0.55, "learning_rate": 0.0008772204433593428, "loss": 1.4932, "step": 5610 }, { "epoch": 0.55, "learning_rate": 0.0008769035654623261, "loss": 1.4502, "step": 5611 }, { "epoch": 0.55, "learning_rate": 0.000876586700115297, "loss": 1.4463, "step": 5612 }, { "epoch": 0.55, "learning_rate": 0.0008762698473505605, "loss": 1.5205, "step": 5613 }, { "epoch": 0.55, "learning_rate": 0.0008759530072004208, "loss": 1.6436, "step": 5614 }, { "epoch": 0.55, "learning_rate": 0.0008756361796971805, "loss": 1.5156, "step": 5615 }, { "epoch": 0.55, "learning_rate": 0.0008753193648731407, "loss": 1.5771, "step": 5616 }, { "epoch": 0.55, "learning_rate": 0.0008750025627606016, "loss": 1.3887, "step": 5617 }, { "epoch": 0.55, "learning_rate": 0.0008746857733918621, "loss": 1.667, "step": 5618 }, { "epoch": 0.55, "learning_rate": 0.0008743689967992197, "loss": 1.4492, "step": 5619 }, { "epoch": 0.55, "learning_rate": 0.0008740522330149702, "loss": 1.4092, "step": 5620 }, { "epoch": 0.55, "learning_rate": 0.0008737354820714087, "loss": 1.6143, "step": 5621 }, { "epoch": 0.55, "learning_rate": 0.0008734187440008287, "loss": 1.5078, "step": 5622 }, { "epoch": 0.55, "learning_rate": 0.0008731020188355223, "loss": 1.5703, "step": 5623 }, { "epoch": 0.55, "learning_rate": 0.0008727853066077806, "loss": 1.4805, "step": 5624 }, { "epoch": 0.55, "learning_rate": 0.0008724686073498932, "loss": 1.6055, "step": 5625 }, { "epoch": 0.55, "learning_rate": 0.0008721519210941484, "loss": 1.6855, "step": 5626 }, { "epoch": 0.55, "learning_rate": 0.0008718352478728331, "loss": 1.3848, "step": 5627 }, { "epoch": 0.55, "learning_rate": 0.0008715185877182328, "loss": 1.4111, "step": 5628 }, { "epoch": 0.55, "learning_rate": 0.000871201940662632, "loss": 1.6826, "step": 5629 }, { "epoch": 0.56, "learning_rate": 0.0008708853067383135, "loss": 1.4941, "step": 5630 }, { "epoch": 0.56, "learning_rate": 0.0008705686859775591, "loss": 1.5732, "step": 5631 }, { "epoch": 0.56, "learning_rate": 0.0008702520784126488, "loss": 1.5566, "step": 5632 }, { "epoch": 0.56, "learning_rate": 0.0008699354840758616, "loss": 1.6006, "step": 5633 }, { "epoch": 0.56, "learning_rate": 0.0008696189029994753, "loss": 1.5352, "step": 5634 }, { "epoch": 0.56, "learning_rate": 0.000869302335215766, "loss": 1.5098, "step": 5635 }, { "epoch": 0.56, "learning_rate": 0.0008689857807570085, "loss": 1.6445, "step": 5636 }, { "epoch": 0.56, "learning_rate": 0.0008686692396554765, "loss": 1.4033, "step": 5637 }, { "epoch": 0.56, "learning_rate": 0.0008683527119434419, "loss": 1.5498, "step": 5638 }, { "epoch": 0.56, "learning_rate": 0.0008680361976531756, "loss": 1.5039, "step": 5639 }, { "epoch": 0.56, "learning_rate": 0.0008677196968169471, "loss": 1.5264, "step": 5640 }, { "epoch": 0.56, "learning_rate": 0.0008674032094670243, "loss": 1.5205, "step": 5641 }, { "epoch": 0.56, "learning_rate": 0.0008670867356356738, "loss": 1.5078, "step": 5642 }, { "epoch": 0.56, "learning_rate": 0.0008667702753551612, "loss": 1.499, "step": 5643 }, { "epoch": 0.56, "learning_rate": 0.0008664538286577504, "loss": 1.5352, "step": 5644 }, { "epoch": 0.56, "learning_rate": 0.0008661373955757036, "loss": 1.5059, "step": 5645 }, { "epoch": 0.56, "learning_rate": 0.0008658209761412822, "loss": 1.2598, "step": 5646 }, { "epoch": 0.56, "learning_rate": 0.0008655045703867458, "loss": 1.4658, "step": 5647 }, { "epoch": 0.56, "learning_rate": 0.0008651881783443529, "loss": 1.459, "step": 5648 }, { "epoch": 0.56, "learning_rate": 0.0008648718000463603, "loss": 1.415, "step": 5649 }, { "epoch": 0.56, "learning_rate": 0.0008645554355250237, "loss": 1.5225, "step": 5650 }, { "epoch": 0.56, "learning_rate": 0.000864239084812597, "loss": 1.498, "step": 5651 }, { "epoch": 0.56, "learning_rate": 0.0008639227479413331, "loss": 1.3711, "step": 5652 }, { "epoch": 0.56, "learning_rate": 0.0008636064249434837, "loss": 1.5527, "step": 5653 }, { "epoch": 0.56, "learning_rate": 0.0008632901158512981, "loss": 1.5801, "step": 5654 }, { "epoch": 0.56, "learning_rate": 0.0008629738206970253, "loss": 1.5713, "step": 5655 }, { "epoch": 0.56, "learning_rate": 0.000862657539512912, "loss": 1.4746, "step": 5656 }, { "epoch": 0.56, "learning_rate": 0.000862341272331204, "loss": 1.5947, "step": 5657 }, { "epoch": 0.56, "learning_rate": 0.0008620250191841455, "loss": 1.5947, "step": 5658 }, { "epoch": 0.56, "learning_rate": 0.0008617087801039793, "loss": 1.5527, "step": 5659 }, { "epoch": 0.56, "learning_rate": 0.0008613925551229466, "loss": 1.6201, "step": 5660 }, { "epoch": 0.56, "learning_rate": 0.0008610763442732877, "loss": 1.5762, "step": 5661 }, { "epoch": 0.56, "learning_rate": 0.000860760147587241, "loss": 1.6191, "step": 5662 }, { "epoch": 0.56, "learning_rate": 0.0008604439650970432, "loss": 1.5273, "step": 5663 }, { "epoch": 0.56, "learning_rate": 0.0008601277968349303, "loss": 1.4375, "step": 5664 }, { "epoch": 0.56, "learning_rate": 0.0008598116428331362, "loss": 1.5146, "step": 5665 }, { "epoch": 0.56, "learning_rate": 0.0008594955031238936, "loss": 1.543, "step": 5666 }, { "epoch": 0.56, "learning_rate": 0.0008591793777394338, "loss": 1.4082, "step": 5667 }, { "epoch": 0.56, "learning_rate": 0.0008588632667119866, "loss": 1.6143, "step": 5668 }, { "epoch": 0.56, "learning_rate": 0.0008585471700737799, "loss": 1.5029, "step": 5669 }, { "epoch": 0.56, "learning_rate": 0.0008582310878570412, "loss": 1.5615, "step": 5670 }, { "epoch": 0.56, "learning_rate": 0.0008579150200939955, "loss": 1.5684, "step": 5671 }, { "epoch": 0.56, "learning_rate": 0.0008575989668168667, "loss": 1.7285, "step": 5672 }, { "epoch": 0.56, "learning_rate": 0.0008572829280578773, "loss": 1.4668, "step": 5673 }, { "epoch": 0.56, "learning_rate": 0.0008569669038492483, "loss": 1.5059, "step": 5674 }, { "epoch": 0.56, "learning_rate": 0.0008566508942231991, "loss": 1.4854, "step": 5675 }, { "epoch": 0.56, "learning_rate": 0.0008563348992119474, "loss": 1.4971, "step": 5676 }, { "epoch": 0.56, "learning_rate": 0.00085601891884771, "loss": 1.5332, "step": 5677 }, { "epoch": 0.56, "learning_rate": 0.0008557029531627017, "loss": 1.4609, "step": 5678 }, { "epoch": 0.56, "learning_rate": 0.0008553870021891361, "loss": 1.54, "step": 5679 }, { "epoch": 0.56, "learning_rate": 0.0008550710659592252, "loss": 1.4541, "step": 5680 }, { "epoch": 0.56, "learning_rate": 0.0008547551445051793, "loss": 1.4688, "step": 5681 }, { "epoch": 0.56, "learning_rate": 0.0008544392378592076, "loss": 1.6162, "step": 5682 }, { "epoch": 0.56, "learning_rate": 0.0008541233460535176, "loss": 1.4639, "step": 5683 }, { "epoch": 0.56, "learning_rate": 0.0008538074691203148, "loss": 1.6445, "step": 5684 }, { "epoch": 0.56, "learning_rate": 0.0008534916070918042, "loss": 1.5039, "step": 5685 }, { "epoch": 0.56, "learning_rate": 0.0008531757600001885, "loss": 1.4014, "step": 5686 }, { "epoch": 0.56, "learning_rate": 0.0008528599278776691, "loss": 1.5781, "step": 5687 }, { "epoch": 0.56, "learning_rate": 0.0008525441107564458, "loss": 1.5713, "step": 5688 }, { "epoch": 0.56, "learning_rate": 0.000852228308668717, "loss": 1.5088, "step": 5689 }, { "epoch": 0.56, "learning_rate": 0.0008519125216466797, "loss": 1.4658, "step": 5690 }, { "epoch": 0.56, "learning_rate": 0.0008515967497225288, "loss": 1.5137, "step": 5691 }, { "epoch": 0.56, "learning_rate": 0.0008512809929284581, "loss": 1.4951, "step": 5692 }, { "epoch": 0.56, "learning_rate": 0.0008509652512966599, "loss": 1.4922, "step": 5693 }, { "epoch": 0.56, "learning_rate": 0.000850649524859325, "loss": 1.4111, "step": 5694 }, { "epoch": 0.56, "learning_rate": 0.0008503338136486424, "loss": 1.5, "step": 5695 }, { "epoch": 0.56, "learning_rate": 0.0008500181176967996, "loss": 1.54, "step": 5696 }, { "epoch": 0.56, "learning_rate": 0.0008497024370359825, "loss": 1.4678, "step": 5697 }, { "epoch": 0.56, "learning_rate": 0.0008493867716983755, "loss": 1.4385, "step": 5698 }, { "epoch": 0.56, "learning_rate": 0.0008490711217161617, "loss": 1.6572, "step": 5699 }, { "epoch": 0.56, "learning_rate": 0.0008487554871215222, "loss": 1.4805, "step": 5700 }, { "epoch": 0.56, "learning_rate": 0.000848439867946637, "loss": 1.3877, "step": 5701 }, { "epoch": 0.56, "learning_rate": 0.0008481242642236836, "loss": 1.4189, "step": 5702 }, { "epoch": 0.56, "learning_rate": 0.0008478086759848393, "loss": 1.4961, "step": 5703 }, { "epoch": 0.56, "learning_rate": 0.000847493103262279, "loss": 1.4707, "step": 5704 }, { "epoch": 0.56, "learning_rate": 0.0008471775460881758, "loss": 1.4434, "step": 5705 }, { "epoch": 0.56, "learning_rate": 0.0008468620044947019, "loss": 1.417, "step": 5706 }, { "epoch": 0.56, "learning_rate": 0.0008465464785140272, "loss": 1.4609, "step": 5707 }, { "epoch": 0.56, "learning_rate": 0.0008462309681783206, "loss": 1.4482, "step": 5708 }, { "epoch": 0.56, "learning_rate": 0.000845915473519749, "loss": 1.5615, "step": 5709 }, { "epoch": 0.56, "learning_rate": 0.000845599994570478, "loss": 1.6143, "step": 5710 }, { "epoch": 0.56, "learning_rate": 0.0008452845313626713, "loss": 1.5283, "step": 5711 }, { "epoch": 0.56, "learning_rate": 0.0008449690839284913, "loss": 1.6025, "step": 5712 }, { "epoch": 0.56, "learning_rate": 0.0008446536523000987, "loss": 1.6113, "step": 5713 }, { "epoch": 0.56, "learning_rate": 0.0008443382365096525, "loss": 1.4131, "step": 5714 }, { "epoch": 0.56, "learning_rate": 0.00084402283658931, "loss": 1.5352, "step": 5715 }, { "epoch": 0.56, "learning_rate": 0.0008437074525712272, "loss": 1.3955, "step": 5716 }, { "epoch": 0.56, "learning_rate": 0.000843392084487558, "loss": 1.5449, "step": 5717 }, { "epoch": 0.56, "learning_rate": 0.0008430767323704552, "loss": 1.4697, "step": 5718 }, { "epoch": 0.56, "learning_rate": 0.0008427613962520697, "loss": 1.583, "step": 5719 }, { "epoch": 0.56, "learning_rate": 0.0008424460761645506, "loss": 1.6084, "step": 5720 }, { "epoch": 0.56, "learning_rate": 0.000842130772140046, "loss": 1.5146, "step": 5721 }, { "epoch": 0.56, "learning_rate": 0.0008418154842107016, "loss": 1.4854, "step": 5722 }, { "epoch": 0.56, "learning_rate": 0.0008415002124086617, "loss": 1.5244, "step": 5723 }, { "epoch": 0.56, "learning_rate": 0.0008411849567660693, "loss": 1.5059, "step": 5724 }, { "epoch": 0.56, "learning_rate": 0.0008408697173150654, "loss": 1.5361, "step": 5725 }, { "epoch": 0.56, "learning_rate": 0.0008405544940877894, "loss": 1.3379, "step": 5726 }, { "epoch": 0.56, "learning_rate": 0.0008402392871163792, "loss": 1.6182, "step": 5727 }, { "epoch": 0.56, "learning_rate": 0.0008399240964329709, "loss": 1.4629, "step": 5728 }, { "epoch": 0.56, "learning_rate": 0.0008396089220696985, "loss": 1.5059, "step": 5729 }, { "epoch": 0.56, "learning_rate": 0.0008392937640586956, "loss": 1.5068, "step": 5730 }, { "epoch": 0.56, "learning_rate": 0.0008389786224320935, "loss": 1.5996, "step": 5731 }, { "epoch": 0.57, "learning_rate": 0.0008386634972220206, "loss": 1.4502, "step": 5732 }, { "epoch": 0.57, "learning_rate": 0.0008383483884606053, "loss": 1.3779, "step": 5733 }, { "epoch": 0.57, "learning_rate": 0.0008380332961799735, "loss": 1.5098, "step": 5734 }, { "epoch": 0.57, "learning_rate": 0.0008377182204122501, "loss": 1.4756, "step": 5735 }, { "epoch": 0.57, "learning_rate": 0.0008374031611895575, "loss": 1.4688, "step": 5736 }, { "epoch": 0.57, "learning_rate": 0.0008370881185440169, "loss": 1.4424, "step": 5737 }, { "epoch": 0.57, "learning_rate": 0.0008367730925077475, "loss": 1.4199, "step": 5738 }, { "epoch": 0.57, "learning_rate": 0.0008364580831128672, "loss": 1.4141, "step": 5739 }, { "epoch": 0.57, "learning_rate": 0.000836143090391492, "loss": 1.5869, "step": 5740 }, { "epoch": 0.57, "learning_rate": 0.000835828114375736, "loss": 1.6357, "step": 5741 }, { "epoch": 0.57, "learning_rate": 0.000835513155097712, "loss": 1.6191, "step": 5742 }, { "epoch": 0.57, "learning_rate": 0.0008351982125895303, "loss": 1.6299, "step": 5743 }, { "epoch": 0.57, "learning_rate": 0.000834883286883301, "loss": 1.5566, "step": 5744 }, { "epoch": 0.57, "learning_rate": 0.000834568378011131, "loss": 1.5195, "step": 5745 }, { "epoch": 0.57, "learning_rate": 0.0008342534860051263, "loss": 1.6318, "step": 5746 }, { "epoch": 0.57, "learning_rate": 0.0008339386108973905, "loss": 1.4775, "step": 5747 }, { "epoch": 0.57, "learning_rate": 0.0008336237527200263, "loss": 1.4951, "step": 5748 }, { "epoch": 0.57, "learning_rate": 0.0008333089115051342, "loss": 1.4092, "step": 5749 }, { "epoch": 0.57, "learning_rate": 0.0008329940872848129, "loss": 1.415, "step": 5750 }, { "epoch": 0.57, "learning_rate": 0.0008326792800911598, "loss": 1.4756, "step": 5751 }, { "epoch": 0.57, "learning_rate": 0.0008323644899562697, "loss": 1.5439, "step": 5752 }, { "epoch": 0.57, "learning_rate": 0.0008320497169122368, "loss": 1.4463, "step": 5753 }, { "epoch": 0.57, "learning_rate": 0.0008317349609911529, "loss": 1.5127, "step": 5754 }, { "epoch": 0.57, "learning_rate": 0.0008314202222251079, "loss": 1.584, "step": 5755 }, { "epoch": 0.57, "learning_rate": 0.0008311055006461905, "loss": 1.4941, "step": 5756 }, { "epoch": 0.57, "learning_rate": 0.0008307907962864871, "loss": 1.4463, "step": 5757 }, { "epoch": 0.57, "learning_rate": 0.0008304761091780827, "loss": 1.6475, "step": 5758 }, { "epoch": 0.57, "learning_rate": 0.0008301614393530605, "loss": 1.5264, "step": 5759 }, { "epoch": 0.57, "learning_rate": 0.0008298467868435018, "loss": 1.4951, "step": 5760 }, { "epoch": 0.57, "learning_rate": 0.0008295321516814859, "loss": 1.5352, "step": 5761 }, { "epoch": 0.57, "learning_rate": 0.0008292175338990912, "loss": 1.4951, "step": 5762 }, { "epoch": 0.57, "learning_rate": 0.0008289029335283935, "loss": 1.4365, "step": 5763 }, { "epoch": 0.57, "learning_rate": 0.0008285883506014671, "loss": 1.5586, "step": 5764 }, { "epoch": 0.57, "learning_rate": 0.0008282737851503845, "loss": 1.418, "step": 5765 }, { "epoch": 0.57, "learning_rate": 0.0008279592372072164, "loss": 1.5488, "step": 5766 }, { "epoch": 0.57, "learning_rate": 0.000827644706804032, "loss": 1.5425, "step": 5767 }, { "epoch": 0.57, "learning_rate": 0.0008273301939728979, "loss": 1.5254, "step": 5768 }, { "epoch": 0.57, "learning_rate": 0.00082701569874588, "loss": 1.4385, "step": 5769 }, { "epoch": 0.57, "learning_rate": 0.0008267012211550414, "loss": 1.4814, "step": 5770 }, { "epoch": 0.57, "learning_rate": 0.0008263867612324444, "loss": 1.5449, "step": 5771 }, { "epoch": 0.57, "learning_rate": 0.0008260723190101487, "loss": 1.3564, "step": 5772 }, { "epoch": 0.57, "learning_rate": 0.0008257578945202124, "loss": 1.5127, "step": 5773 }, { "epoch": 0.57, "learning_rate": 0.0008254434877946921, "loss": 1.4316, "step": 5774 }, { "epoch": 0.57, "learning_rate": 0.0008251290988656421, "loss": 1.4854, "step": 5775 }, { "epoch": 0.57, "learning_rate": 0.0008248147277651152, "loss": 1.5762, "step": 5776 }, { "epoch": 0.57, "learning_rate": 0.0008245003745251624, "loss": 1.4844, "step": 5777 }, { "epoch": 0.57, "learning_rate": 0.0008241860391778327, "loss": 1.6279, "step": 5778 }, { "epoch": 0.57, "learning_rate": 0.0008238717217551732, "loss": 1.4521, "step": 5779 }, { "epoch": 0.57, "learning_rate": 0.0008235574222892298, "loss": 1.4404, "step": 5780 }, { "epoch": 0.57, "learning_rate": 0.0008232431408120458, "loss": 1.4814, "step": 5781 }, { "epoch": 0.57, "learning_rate": 0.0008229288773556631, "loss": 1.541, "step": 5782 }, { "epoch": 0.57, "learning_rate": 0.0008226146319521214, "loss": 1.6895, "step": 5783 }, { "epoch": 0.57, "learning_rate": 0.0008223004046334591, "loss": 1.4521, "step": 5784 }, { "epoch": 0.57, "learning_rate": 0.0008219861954317123, "loss": 1.6221, "step": 5785 }, { "epoch": 0.57, "learning_rate": 0.0008216720043789154, "loss": 1.5039, "step": 5786 }, { "epoch": 0.57, "learning_rate": 0.0008213578315071013, "loss": 1.6309, "step": 5787 }, { "epoch": 0.57, "learning_rate": 0.0008210436768483002, "loss": 1.4141, "step": 5788 }, { "epoch": 0.57, "learning_rate": 0.000820729540434541, "loss": 1.5088, "step": 5789 }, { "epoch": 0.57, "learning_rate": 0.000820415422297851, "loss": 1.5195, "step": 5790 }, { "epoch": 0.57, "learning_rate": 0.0008201013224702552, "loss": 1.6836, "step": 5791 }, { "epoch": 0.57, "learning_rate": 0.0008197872409837767, "loss": 1.4492, "step": 5792 }, { "epoch": 0.57, "learning_rate": 0.0008194731778704367, "loss": 1.3877, "step": 5793 }, { "epoch": 0.57, "learning_rate": 0.0008191591331622554, "loss": 1.6152, "step": 5794 }, { "epoch": 0.57, "learning_rate": 0.0008188451068912501, "loss": 1.5488, "step": 5795 }, { "epoch": 0.57, "learning_rate": 0.0008185310990894365, "loss": 1.4004, "step": 5796 }, { "epoch": 0.57, "learning_rate": 0.0008182171097888284, "loss": 1.5977, "step": 5797 }, { "epoch": 0.57, "learning_rate": 0.000817903139021438, "loss": 1.457, "step": 5798 }, { "epoch": 0.57, "learning_rate": 0.0008175891868192753, "loss": 1.4473, "step": 5799 }, { "epoch": 0.57, "learning_rate": 0.0008172752532143484, "loss": 1.5645, "step": 5800 }, { "epoch": 0.57, "learning_rate": 0.0008169613382386638, "loss": 1.5029, "step": 5801 }, { "epoch": 0.57, "learning_rate": 0.0008166474419242256, "loss": 1.5527, "step": 5802 }, { "epoch": 0.57, "learning_rate": 0.0008163335643030367, "loss": 1.5205, "step": 5803 }, { "epoch": 0.57, "learning_rate": 0.0008160197054070975, "loss": 1.4268, "step": 5804 }, { "epoch": 0.57, "learning_rate": 0.0008157058652684068, "loss": 1.459, "step": 5805 }, { "epoch": 0.57, "learning_rate": 0.0008153920439189613, "loss": 1.46, "step": 5806 }, { "epoch": 0.57, "learning_rate": 0.0008150782413907559, "loss": 1.5879, "step": 5807 }, { "epoch": 0.57, "learning_rate": 0.0008147644577157835, "loss": 1.3965, "step": 5808 }, { "epoch": 0.57, "learning_rate": 0.0008144506929260351, "loss": 1.4922, "step": 5809 }, { "epoch": 0.57, "learning_rate": 0.0008141369470535, "loss": 1.4873, "step": 5810 }, { "epoch": 0.57, "learning_rate": 0.0008138232201301649, "loss": 1.5029, "step": 5811 }, { "epoch": 0.57, "learning_rate": 0.0008135095121880157, "loss": 1.5303, "step": 5812 }, { "epoch": 0.57, "learning_rate": 0.0008131958232590352, "loss": 1.5898, "step": 5813 }, { "epoch": 0.57, "learning_rate": 0.000812882153375205, "loss": 1.4038, "step": 5814 }, { "epoch": 0.57, "learning_rate": 0.0008125685025685044, "loss": 1.5293, "step": 5815 }, { "epoch": 0.57, "learning_rate": 0.0008122548708709111, "loss": 1.5605, "step": 5816 }, { "epoch": 0.57, "learning_rate": 0.0008119412583144003, "loss": 1.4609, "step": 5817 }, { "epoch": 0.57, "learning_rate": 0.000811627664930946, "loss": 1.5703, "step": 5818 }, { "epoch": 0.57, "learning_rate": 0.0008113140907525194, "loss": 1.5439, "step": 5819 }, { "epoch": 0.57, "learning_rate": 0.0008110005358110901, "loss": 1.4248, "step": 5820 }, { "epoch": 0.57, "learning_rate": 0.0008106870001386262, "loss": 1.6396, "step": 5821 }, { "epoch": 0.57, "learning_rate": 0.0008103734837670935, "loss": 1.4893, "step": 5822 }, { "epoch": 0.57, "learning_rate": 0.0008100599867284555, "loss": 1.5146, "step": 5823 }, { "epoch": 0.57, "learning_rate": 0.000809746509054674, "loss": 1.5156, "step": 5824 }, { "epoch": 0.57, "learning_rate": 0.0008094330507777089, "loss": 1.6279, "step": 5825 }, { "epoch": 0.57, "learning_rate": 0.0008091196119295181, "loss": 1.417, "step": 5826 }, { "epoch": 0.57, "learning_rate": 0.0008088061925420573, "loss": 1.501, "step": 5827 }, { "epoch": 0.57, "learning_rate": 0.0008084927926472807, "loss": 1.5215, "step": 5828 }, { "epoch": 0.57, "learning_rate": 0.0008081794122771396, "loss": 1.6846, "step": 5829 }, { "epoch": 0.57, "learning_rate": 0.0008078660514635848, "loss": 1.6006, "step": 5830 }, { "epoch": 0.57, "learning_rate": 0.0008075527102385634, "loss": 1.5703, "step": 5831 }, { "epoch": 0.57, "learning_rate": 0.0008072393886340217, "loss": 1.4658, "step": 5832 }, { "epoch": 0.58, "learning_rate": 0.0008069260866819038, "loss": 1.5156, "step": 5833 }, { "epoch": 0.58, "learning_rate": 0.000806612804414151, "loss": 1.4409, "step": 5834 }, { "epoch": 0.58, "learning_rate": 0.0008062995418627038, "loss": 1.4912, "step": 5835 }, { "epoch": 0.58, "learning_rate": 0.0008059862990594998, "loss": 1.5518, "step": 5836 }, { "epoch": 0.58, "learning_rate": 0.000805673076036475, "loss": 1.5645, "step": 5837 }, { "epoch": 0.58, "learning_rate": 0.000805359872825563, "loss": 1.5215, "step": 5838 }, { "epoch": 0.58, "learning_rate": 0.0008050466894586961, "loss": 1.5146, "step": 5839 }, { "epoch": 0.58, "learning_rate": 0.0008047335259678038, "loss": 1.5439, "step": 5840 }, { "epoch": 0.58, "learning_rate": 0.0008044203823848139, "loss": 1.459, "step": 5841 }, { "epoch": 0.58, "learning_rate": 0.0008041072587416524, "loss": 1.4453, "step": 5842 }, { "epoch": 0.58, "learning_rate": 0.0008037941550702432, "loss": 1.5068, "step": 5843 }, { "epoch": 0.58, "learning_rate": 0.0008034810714025074, "loss": 1.5273, "step": 5844 }, { "epoch": 0.58, "learning_rate": 0.0008031680077703648, "loss": 1.5439, "step": 5845 }, { "epoch": 0.58, "learning_rate": 0.0008028549642057332, "loss": 1.5654, "step": 5846 }, { "epoch": 0.58, "learning_rate": 0.0008025419407405283, "loss": 1.3516, "step": 5847 }, { "epoch": 0.58, "learning_rate": 0.0008022289374066632, "loss": 1.4033, "step": 5848 }, { "epoch": 0.58, "learning_rate": 0.0008019159542360498, "loss": 1.5137, "step": 5849 }, { "epoch": 0.58, "learning_rate": 0.0008016029912605974, "loss": 1.5723, "step": 5850 }, { "epoch": 0.58, "learning_rate": 0.000801290048512213, "loss": 1.5957, "step": 5851 }, { "epoch": 0.58, "learning_rate": 0.000800977126022802, "loss": 1.543, "step": 5852 }, { "epoch": 0.58, "learning_rate": 0.0008006642238242682, "loss": 1.4688, "step": 5853 }, { "epoch": 0.58, "learning_rate": 0.0008003513419485122, "loss": 1.5732, "step": 5854 }, { "epoch": 0.58, "learning_rate": 0.0008000384804274333, "loss": 1.542, "step": 5855 }, { "epoch": 0.58, "learning_rate": 0.0007997256392929285, "loss": 1.5723, "step": 5856 }, { "epoch": 0.58, "learning_rate": 0.0007994128185768927, "loss": 1.4443, "step": 5857 }, { "epoch": 0.58, "learning_rate": 0.0007991000183112186, "loss": 1.5137, "step": 5858 }, { "epoch": 0.58, "learning_rate": 0.0007987872385277973, "loss": 1.4893, "step": 5859 }, { "epoch": 0.58, "learning_rate": 0.0007984744792585171, "loss": 1.5195, "step": 5860 }, { "epoch": 0.58, "learning_rate": 0.0007981617405352644, "loss": 1.4873, "step": 5861 }, { "epoch": 0.58, "learning_rate": 0.0007978490223899245, "loss": 1.5342, "step": 5862 }, { "epoch": 0.58, "learning_rate": 0.0007975363248543792, "loss": 1.624, "step": 5863 }, { "epoch": 0.58, "learning_rate": 0.000797223647960509, "loss": 1.4775, "step": 5864 }, { "epoch": 0.58, "learning_rate": 0.0007969109917401921, "loss": 1.5361, "step": 5865 }, { "epoch": 0.58, "learning_rate": 0.0007965983562253044, "loss": 1.6387, "step": 5866 }, { "epoch": 0.58, "learning_rate": 0.0007962857414477199, "loss": 1.4209, "step": 5867 }, { "epoch": 0.58, "learning_rate": 0.0007959731474393106, "loss": 1.4209, "step": 5868 }, { "epoch": 0.58, "learning_rate": 0.0007956605742319463, "loss": 1.4863, "step": 5869 }, { "epoch": 0.58, "learning_rate": 0.000795348021857494, "loss": 1.584, "step": 5870 }, { "epoch": 0.58, "learning_rate": 0.00079503549034782, "loss": 1.5303, "step": 5871 }, { "epoch": 0.58, "learning_rate": 0.0007947229797347872, "loss": 1.4941, "step": 5872 }, { "epoch": 0.58, "learning_rate": 0.0007944104900502572, "loss": 1.4492, "step": 5873 }, { "epoch": 0.58, "learning_rate": 0.0007940980213260887, "loss": 1.5029, "step": 5874 }, { "epoch": 0.58, "learning_rate": 0.0007937855735941389, "loss": 1.5664, "step": 5875 }, { "epoch": 0.58, "learning_rate": 0.0007934731468862624, "loss": 1.4307, "step": 5876 }, { "epoch": 0.58, "learning_rate": 0.0007931607412343121, "loss": 1.4678, "step": 5877 }, { "epoch": 0.58, "learning_rate": 0.0007928483566701386, "loss": 1.502, "step": 5878 }, { "epoch": 0.58, "learning_rate": 0.0007925359932255898, "loss": 1.4551, "step": 5879 }, { "epoch": 0.58, "learning_rate": 0.0007922236509325124, "loss": 1.4424, "step": 5880 }, { "epoch": 0.58, "learning_rate": 0.0007919113298227505, "loss": 1.4434, "step": 5881 }, { "epoch": 0.58, "learning_rate": 0.0007915990299281457, "loss": 1.5869, "step": 5882 }, { "epoch": 0.58, "learning_rate": 0.0007912867512805381, "loss": 1.6641, "step": 5883 }, { "epoch": 0.58, "learning_rate": 0.000790974493911765, "loss": 1.5176, "step": 5884 }, { "epoch": 0.58, "learning_rate": 0.0007906622578536618, "loss": 1.7227, "step": 5885 }, { "epoch": 0.58, "learning_rate": 0.000790350043138062, "loss": 1.46, "step": 5886 }, { "epoch": 0.58, "learning_rate": 0.0007900378497967965, "loss": 1.4512, "step": 5887 }, { "epoch": 0.58, "learning_rate": 0.0007897256778616939, "loss": 1.6211, "step": 5888 }, { "epoch": 0.58, "learning_rate": 0.0007894135273645815, "loss": 1.4141, "step": 5889 }, { "epoch": 0.58, "learning_rate": 0.0007891013983372836, "loss": 1.5068, "step": 5890 }, { "epoch": 0.58, "learning_rate": 0.0007887892908116223, "loss": 1.458, "step": 5891 }, { "epoch": 0.58, "learning_rate": 0.0007884772048194179, "loss": 1.4766, "step": 5892 }, { "epoch": 0.58, "learning_rate": 0.0007881651403924885, "loss": 1.4814, "step": 5893 }, { "epoch": 0.58, "learning_rate": 0.0007878530975626496, "loss": 1.5234, "step": 5894 }, { "epoch": 0.58, "learning_rate": 0.0007875410763617148, "loss": 1.5342, "step": 5895 }, { "epoch": 0.58, "learning_rate": 0.0007872290768214955, "loss": 1.5684, "step": 5896 }, { "epoch": 0.58, "learning_rate": 0.0007869170989738008, "loss": 1.5195, "step": 5897 }, { "epoch": 0.58, "learning_rate": 0.0007866051428504374, "loss": 1.5537, "step": 5898 }, { "epoch": 0.58, "learning_rate": 0.0007862932084832104, "loss": 1.4697, "step": 5899 }, { "epoch": 0.58, "learning_rate": 0.0007859812959039225, "loss": 1.374, "step": 5900 }, { "epoch": 0.58, "learning_rate": 0.0007856694051443731, "loss": 1.5889, "step": 5901 }, { "epoch": 0.58, "learning_rate": 0.0007853575362363609, "loss": 1.4551, "step": 5902 }, { "epoch": 0.58, "learning_rate": 0.0007850456892116811, "loss": 1.5771, "step": 5903 }, { "epoch": 0.58, "learning_rate": 0.000784733864102128, "loss": 1.5967, "step": 5904 }, { "epoch": 0.58, "learning_rate": 0.0007844220609394925, "loss": 1.5234, "step": 5905 }, { "epoch": 0.58, "learning_rate": 0.000784110279755564, "loss": 1.5166, "step": 5906 }, { "epoch": 0.58, "learning_rate": 0.0007837985205821291, "loss": 1.5479, "step": 5907 }, { "epoch": 0.58, "learning_rate": 0.0007834867834509726, "loss": 1.5303, "step": 5908 }, { "epoch": 0.58, "learning_rate": 0.0007831750683938767, "loss": 1.6299, "step": 5909 }, { "epoch": 0.58, "learning_rate": 0.0007828633754426217, "loss": 1.4531, "step": 5910 }, { "epoch": 0.58, "learning_rate": 0.0007825517046289854, "loss": 1.6348, "step": 5911 }, { "epoch": 0.58, "learning_rate": 0.0007822400559847432, "loss": 1.5205, "step": 5912 }, { "epoch": 0.58, "learning_rate": 0.0007819284295416688, "loss": 1.5156, "step": 5913 }, { "epoch": 0.58, "learning_rate": 0.0007816168253315333, "loss": 1.6201, "step": 5914 }, { "epoch": 0.58, "learning_rate": 0.0007813052433861055, "loss": 1.5654, "step": 5915 }, { "epoch": 0.58, "learning_rate": 0.0007809936837371515, "loss": 1.4326, "step": 5916 }, { "epoch": 0.58, "learning_rate": 0.000780682146416436, "loss": 1.6035, "step": 5917 }, { "epoch": 0.58, "learning_rate": 0.0007803706314557211, "loss": 1.4414, "step": 5918 }, { "epoch": 0.58, "learning_rate": 0.0007800591388867663, "loss": 1.4971, "step": 5919 }, { "epoch": 0.58, "learning_rate": 0.0007797476687413289, "loss": 1.5049, "step": 5920 }, { "epoch": 0.58, "learning_rate": 0.0007794362210511643, "loss": 1.5576, "step": 5921 }, { "epoch": 0.58, "learning_rate": 0.0007791247958480255, "loss": 1.5098, "step": 5922 }, { "epoch": 0.58, "learning_rate": 0.0007788133931636627, "loss": 1.4268, "step": 5923 }, { "epoch": 0.58, "learning_rate": 0.0007785020130298244, "loss": 1.3867, "step": 5924 }, { "epoch": 0.58, "learning_rate": 0.0007781906554782566, "loss": 1.5156, "step": 5925 }, { "epoch": 0.58, "learning_rate": 0.000777879320540703, "loss": 1.4512, "step": 5926 }, { "epoch": 0.58, "learning_rate": 0.0007775680082489047, "loss": 1.4316, "step": 5927 }, { "epoch": 0.58, "learning_rate": 0.0007772567186346011, "loss": 1.437, "step": 5928 }, { "epoch": 0.58, "learning_rate": 0.0007769454517295288, "loss": 1.5039, "step": 5929 }, { "epoch": 0.58, "learning_rate": 0.000776634207565422, "loss": 1.418, "step": 5930 }, { "epoch": 0.58, "learning_rate": 0.0007763229861740132, "loss": 1.3711, "step": 5931 }, { "epoch": 0.58, "learning_rate": 0.0007760117875870322, "loss": 1.4082, "step": 5932 }, { "epoch": 0.58, "learning_rate": 0.0007757006118362062, "loss": 1.5112, "step": 5933 }, { "epoch": 0.58, "learning_rate": 0.0007753894589532604, "loss": 1.5938, "step": 5934 }, { "epoch": 0.59, "learning_rate": 0.0007750783289699178, "loss": 1.4844, "step": 5935 }, { "epoch": 0.59, "learning_rate": 0.0007747672219178988, "loss": 1.5137, "step": 5936 }, { "epoch": 0.59, "learning_rate": 0.0007744561378289212, "loss": 1.4688, "step": 5937 }, { "epoch": 0.59, "learning_rate": 0.0007741450767347012, "loss": 1.3965, "step": 5938 }, { "epoch": 0.59, "learning_rate": 0.0007738340386669521, "loss": 1.6084, "step": 5939 }, { "epoch": 0.59, "learning_rate": 0.000773523023657385, "loss": 1.4678, "step": 5940 }, { "epoch": 0.59, "learning_rate": 0.0007732120317377089, "loss": 1.4902, "step": 5941 }, { "epoch": 0.59, "learning_rate": 0.0007729010629396298, "loss": 1.5791, "step": 5942 }, { "epoch": 0.59, "learning_rate": 0.0007725901172948521, "loss": 1.4346, "step": 5943 }, { "epoch": 0.59, "learning_rate": 0.0007722791948350772, "loss": 1.3457, "step": 5944 }, { "epoch": 0.59, "learning_rate": 0.0007719682955920046, "loss": 1.4893, "step": 5945 }, { "epoch": 0.59, "learning_rate": 0.0007716574195973311, "loss": 1.54, "step": 5946 }, { "epoch": 0.59, "learning_rate": 0.0007713465668827514, "loss": 1.5215, "step": 5947 }, { "epoch": 0.59, "learning_rate": 0.0007710357374799575, "loss": 1.4717, "step": 5948 }, { "epoch": 0.59, "learning_rate": 0.0007707249314206395, "loss": 1.5479, "step": 5949 }, { "epoch": 0.59, "learning_rate": 0.0007704141487364847, "loss": 1.6328, "step": 5950 }, { "epoch": 0.59, "learning_rate": 0.0007701033894591783, "loss": 1.5283, "step": 5951 }, { "epoch": 0.59, "learning_rate": 0.000769792653620403, "loss": 1.5469, "step": 5952 }, { "epoch": 0.59, "learning_rate": 0.0007694819412518389, "loss": 1.5283, "step": 5953 }, { "epoch": 0.59, "learning_rate": 0.0007691712523851639, "loss": 1.5566, "step": 5954 }, { "epoch": 0.59, "learning_rate": 0.0007688605870520538, "loss": 1.5537, "step": 5955 }, { "epoch": 0.59, "learning_rate": 0.0007685499452841815, "loss": 1.5928, "step": 5956 }, { "epoch": 0.59, "learning_rate": 0.0007682393271132176, "loss": 1.3252, "step": 5957 }, { "epoch": 0.59, "learning_rate": 0.0007679287325708304, "loss": 1.4141, "step": 5958 }, { "epoch": 0.59, "learning_rate": 0.000767618161688686, "loss": 1.4824, "step": 5959 }, { "epoch": 0.59, "learning_rate": 0.0007673076144984476, "loss": 1.5332, "step": 5960 }, { "epoch": 0.59, "learning_rate": 0.0007669970910317765, "loss": 1.4414, "step": 5961 }, { "epoch": 0.59, "learning_rate": 0.000766686591320331, "loss": 1.6143, "step": 5962 }, { "epoch": 0.59, "learning_rate": 0.0007663761153957676, "loss": 1.5303, "step": 5963 }, { "epoch": 0.59, "learning_rate": 0.0007660656632897403, "loss": 1.5449, "step": 5964 }, { "epoch": 0.59, "learning_rate": 0.0007657552350339002, "loss": 1.4912, "step": 5965 }, { "epoch": 0.59, "learning_rate": 0.0007654448306598961, "loss": 1.5205, "step": 5966 }, { "epoch": 0.59, "learning_rate": 0.0007651344501993747, "loss": 1.5078, "step": 5967 }, { "epoch": 0.59, "learning_rate": 0.0007648240936839801, "loss": 1.4346, "step": 5968 }, { "epoch": 0.59, "learning_rate": 0.0007645137611453538, "loss": 1.4004, "step": 5969 }, { "epoch": 0.59, "learning_rate": 0.0007642034526151349, "loss": 1.4824, "step": 5970 }, { "epoch": 0.59, "learning_rate": 0.0007638931681249601, "loss": 1.6182, "step": 5971 }, { "epoch": 0.59, "learning_rate": 0.000763582907706464, "loss": 1.4141, "step": 5972 }, { "epoch": 0.59, "learning_rate": 0.0007632726713912782, "loss": 1.5576, "step": 5973 }, { "epoch": 0.59, "learning_rate": 0.0007629624592110321, "loss": 1.501, "step": 5974 }, { "epoch": 0.59, "learning_rate": 0.0007626522711973526, "loss": 1.3975, "step": 5975 }, { "epoch": 0.59, "learning_rate": 0.0007623421073818641, "loss": 1.5264, "step": 5976 }, { "epoch": 0.59, "learning_rate": 0.0007620319677961886, "loss": 1.5479, "step": 5977 }, { "epoch": 0.59, "learning_rate": 0.0007617218524719456, "loss": 1.376, "step": 5978 }, { "epoch": 0.59, "learning_rate": 0.0007614117614407521, "loss": 1.5732, "step": 5979 }, { "epoch": 0.59, "learning_rate": 0.0007611016947342225, "loss": 1.5811, "step": 5980 }, { "epoch": 0.59, "learning_rate": 0.0007607916523839692, "loss": 1.4785, "step": 5981 }, { "epoch": 0.59, "learning_rate": 0.0007604816344216018, "loss": 1.5801, "step": 5982 }, { "epoch": 0.59, "learning_rate": 0.0007601716408787272, "loss": 1.4346, "step": 5983 }, { "epoch": 0.59, "learning_rate": 0.00075986167178695, "loss": 1.457, "step": 5984 }, { "epoch": 0.59, "learning_rate": 0.0007595517271778724, "loss": 1.5166, "step": 5985 }, { "epoch": 0.59, "learning_rate": 0.0007592418070830938, "loss": 1.5068, "step": 5986 }, { "epoch": 0.59, "learning_rate": 0.0007589319115342118, "loss": 1.5254, "step": 5987 }, { "epoch": 0.59, "learning_rate": 0.0007586220405628206, "loss": 1.3926, "step": 5988 }, { "epoch": 0.59, "learning_rate": 0.0007583121942005124, "loss": 1.5576, "step": 5989 }, { "epoch": 0.59, "learning_rate": 0.000758002372478877, "loss": 1.6406, "step": 5990 }, { "epoch": 0.59, "learning_rate": 0.0007576925754295012, "loss": 1.3818, "step": 5991 }, { "epoch": 0.59, "learning_rate": 0.0007573828030839699, "loss": 1.498, "step": 5992 }, { "epoch": 0.59, "learning_rate": 0.000757073055473865, "loss": 1.4824, "step": 5993 }, { "epoch": 0.59, "learning_rate": 0.000756763332630766, "loss": 1.4785, "step": 5994 }, { "epoch": 0.59, "learning_rate": 0.0007564536345862499, "loss": 1.4619, "step": 5995 }, { "epoch": 0.59, "learning_rate": 0.0007561439613718913, "loss": 1.5273, "step": 5996 }, { "epoch": 0.59, "learning_rate": 0.0007558343130192622, "loss": 1.5801, "step": 5997 }, { "epoch": 0.59, "learning_rate": 0.0007555246895599314, "loss": 1.3994, "step": 5998 }, { "epoch": 0.59, "learning_rate": 0.0007552150910254668, "loss": 1.585, "step": 5999 }, { "epoch": 0.59, "learning_rate": 0.0007549055174474322, "loss": 1.4893, "step": 6000 }, { "epoch": 0.59, "learning_rate": 0.0007545959688573893, "loss": 1.5947, "step": 6001 }, { "epoch": 0.59, "learning_rate": 0.0007542864452868976, "loss": 1.501, "step": 6002 }, { "epoch": 0.59, "learning_rate": 0.0007539769467675136, "loss": 1.4121, "step": 6003 }, { "epoch": 0.59, "learning_rate": 0.0007536674733307915, "loss": 1.4277, "step": 6004 }, { "epoch": 0.59, "learning_rate": 0.000753358025008283, "loss": 1.4844, "step": 6005 }, { "epoch": 0.59, "learning_rate": 0.0007530486018315371, "loss": 1.5166, "step": 6006 }, { "epoch": 0.59, "learning_rate": 0.0007527392038320998, "loss": 1.6035, "step": 6007 }, { "epoch": 0.59, "learning_rate": 0.0007524298310415156, "loss": 1.5537, "step": 6008 }, { "epoch": 0.59, "learning_rate": 0.0007521204834913258, "loss": 1.4824, "step": 6009 }, { "epoch": 0.59, "learning_rate": 0.0007518111612130688, "loss": 1.5527, "step": 6010 }, { "epoch": 0.59, "learning_rate": 0.000751501864238281, "loss": 1.3984, "step": 6011 }, { "epoch": 0.59, "learning_rate": 0.0007511925925984961, "loss": 1.5557, "step": 6012 }, { "epoch": 0.59, "learning_rate": 0.0007508833463252449, "loss": 1.4971, "step": 6013 }, { "epoch": 0.59, "learning_rate": 0.0007505741254500558, "loss": 1.4004, "step": 6014 }, { "epoch": 0.59, "learning_rate": 0.0007502649300044546, "loss": 1.4619, "step": 6015 }, { "epoch": 0.59, "learning_rate": 0.0007499557600199647, "loss": 1.5117, "step": 6016 }, { "epoch": 0.59, "learning_rate": 0.0007496466155281068, "loss": 1.6143, "step": 6017 }, { "epoch": 0.59, "learning_rate": 0.0007493374965603987, "loss": 1.5225, "step": 6018 }, { "epoch": 0.59, "learning_rate": 0.0007490284031483561, "loss": 1.5996, "step": 6019 }, { "epoch": 0.59, "learning_rate": 0.0007487193353234918, "loss": 1.4053, "step": 6020 }, { "epoch": 0.59, "learning_rate": 0.0007484102931173157, "loss": 1.4473, "step": 6021 }, { "epoch": 0.59, "learning_rate": 0.0007481012765613359, "loss": 1.4844, "step": 6022 }, { "epoch": 0.59, "learning_rate": 0.0007477922856870572, "loss": 1.5693, "step": 6023 }, { "epoch": 0.59, "learning_rate": 0.0007474833205259821, "loss": 1.4209, "step": 6024 }, { "epoch": 0.59, "learning_rate": 0.0007471743811096101, "loss": 1.5801, "step": 6025 }, { "epoch": 0.59, "learning_rate": 0.0007468654674694386, "loss": 1.4648, "step": 6026 }, { "epoch": 0.59, "learning_rate": 0.000746556579636962, "loss": 1.5947, "step": 6027 }, { "epoch": 0.59, "learning_rate": 0.0007462477176436721, "loss": 1.4434, "step": 6028 }, { "epoch": 0.59, "learning_rate": 0.0007459388815210583, "loss": 1.4102, "step": 6029 }, { "epoch": 0.59, "learning_rate": 0.0007456300713006071, "loss": 1.625, "step": 6030 }, { "epoch": 0.59, "learning_rate": 0.0007453212870138027, "loss": 1.5225, "step": 6031 }, { "epoch": 0.59, "learning_rate": 0.0007450125286921261, "loss": 1.5674, "step": 6032 }, { "epoch": 0.59, "learning_rate": 0.0007447037963670562, "loss": 1.4785, "step": 6033 }, { "epoch": 0.59, "learning_rate": 0.0007443950900700689, "loss": 1.417, "step": 6034 }, { "epoch": 0.59, "learning_rate": 0.0007440864098326377, "loss": 1.5049, "step": 6035 }, { "epoch": 0.6, "learning_rate": 0.0007437777556862333, "loss": 1.666, "step": 6036 }, { "epoch": 0.6, "learning_rate": 0.0007434691276623236, "loss": 1.4551, "step": 6037 }, { "epoch": 0.6, "learning_rate": 0.0007431605257923741, "loss": 1.4668, "step": 6038 }, { "epoch": 0.6, "learning_rate": 0.0007428519501078474, "loss": 1.4268, "step": 6039 }, { "epoch": 0.6, "learning_rate": 0.0007425434006402039, "loss": 1.5732, "step": 6040 }, { "epoch": 0.6, "learning_rate": 0.0007422348774209007, "loss": 1.4551, "step": 6041 }, { "epoch": 0.6, "learning_rate": 0.0007419263804813926, "loss": 1.5342, "step": 6042 }, { "epoch": 0.6, "learning_rate": 0.0007416179098531317, "loss": 1.5322, "step": 6043 }, { "epoch": 0.6, "learning_rate": 0.0007413094655675671, "loss": 1.5957, "step": 6044 }, { "epoch": 0.6, "learning_rate": 0.0007410010476561458, "loss": 1.4141, "step": 6045 }, { "epoch": 0.6, "learning_rate": 0.0007406926561503114, "loss": 1.4072, "step": 6046 }, { "epoch": 0.6, "learning_rate": 0.0007403842910815055, "loss": 1.4834, "step": 6047 }, { "epoch": 0.6, "learning_rate": 0.0007400759524811664, "loss": 1.5, "step": 6048 }, { "epoch": 0.6, "learning_rate": 0.0007397676403807303, "loss": 1.4922, "step": 6049 }, { "epoch": 0.6, "learning_rate": 0.0007394593548116302, "loss": 1.5459, "step": 6050 }, { "epoch": 0.6, "learning_rate": 0.0007391510958052966, "loss": 1.5547, "step": 6051 }, { "epoch": 0.6, "learning_rate": 0.0007388428633931573, "loss": 1.4932, "step": 6052 }, { "epoch": 0.6, "learning_rate": 0.0007385346576066374, "loss": 1.4209, "step": 6053 }, { "epoch": 0.6, "learning_rate": 0.0007382264784771592, "loss": 1.4355, "step": 6054 }, { "epoch": 0.6, "learning_rate": 0.0007379183260361422, "loss": 1.5059, "step": 6055 }, { "epoch": 0.6, "learning_rate": 0.0007376102003150036, "loss": 1.5195, "step": 6056 }, { "epoch": 0.6, "learning_rate": 0.0007373021013451571, "loss": 1.5293, "step": 6057 }, { "epoch": 0.6, "learning_rate": 0.0007369940291580146, "loss": 1.5801, "step": 6058 }, { "epoch": 0.6, "learning_rate": 0.0007366859837849848, "loss": 1.5576, "step": 6059 }, { "epoch": 0.6, "learning_rate": 0.0007363779652574735, "loss": 1.542, "step": 6060 }, { "epoch": 0.6, "learning_rate": 0.0007360699736068841, "loss": 1.5137, "step": 6061 }, { "epoch": 0.6, "learning_rate": 0.0007357620088646171, "loss": 1.6338, "step": 6062 }, { "epoch": 0.6, "learning_rate": 0.0007354540710620702, "loss": 1.4424, "step": 6063 }, { "epoch": 0.6, "learning_rate": 0.0007351461602306384, "loss": 1.4355, "step": 6064 }, { "epoch": 0.6, "learning_rate": 0.000734838276401714, "loss": 1.4395, "step": 6065 }, { "epoch": 0.6, "learning_rate": 0.0007345304196066865, "loss": 1.5078, "step": 6066 }, { "epoch": 0.6, "learning_rate": 0.0007342225898769428, "loss": 1.3848, "step": 6067 }, { "epoch": 0.6, "learning_rate": 0.0007339147872438672, "loss": 1.5625, "step": 6068 }, { "epoch": 0.6, "learning_rate": 0.0007336070117388401, "loss": 1.4443, "step": 6069 }, { "epoch": 0.6, "learning_rate": 0.0007332992633932405, "loss": 1.4512, "step": 6070 }, { "epoch": 0.6, "learning_rate": 0.0007329915422384439, "loss": 1.4551, "step": 6071 }, { "epoch": 0.6, "learning_rate": 0.0007326838483058236, "loss": 1.5127, "step": 6072 }, { "epoch": 0.6, "learning_rate": 0.0007323761816267494, "loss": 1.5898, "step": 6073 }, { "epoch": 0.6, "learning_rate": 0.000732068542232589, "loss": 1.458, "step": 6074 }, { "epoch": 0.6, "learning_rate": 0.0007317609301547068, "loss": 1.4668, "step": 6075 }, { "epoch": 0.6, "learning_rate": 0.0007314533454244645, "loss": 1.3789, "step": 6076 }, { "epoch": 0.6, "learning_rate": 0.0007311457880732214, "loss": 1.4424, "step": 6077 }, { "epoch": 0.6, "learning_rate": 0.0007308382581323336, "loss": 1.4775, "step": 6078 }, { "epoch": 0.6, "learning_rate": 0.0007305307556331545, "loss": 1.4463, "step": 6079 }, { "epoch": 0.6, "learning_rate": 0.0007302232806070346, "loss": 1.541, "step": 6080 }, { "epoch": 0.6, "learning_rate": 0.0007299158330853221, "loss": 1.5898, "step": 6081 }, { "epoch": 0.6, "learning_rate": 0.0007296084130993621, "loss": 1.4268, "step": 6082 }, { "epoch": 0.6, "learning_rate": 0.0007293010206804964, "loss": 1.4551, "step": 6083 }, { "epoch": 0.6, "learning_rate": 0.0007289936558600648, "loss": 1.4717, "step": 6084 }, { "epoch": 0.6, "learning_rate": 0.0007286863186694035, "loss": 1.4863, "step": 6085 }, { "epoch": 0.6, "learning_rate": 0.0007283790091398468, "loss": 1.5059, "step": 6086 }, { "epoch": 0.6, "learning_rate": 0.000728071727302725, "loss": 1.502, "step": 6087 }, { "epoch": 0.6, "learning_rate": 0.0007277644731893671, "loss": 1.4062, "step": 6088 }, { "epoch": 0.6, "learning_rate": 0.0007274572468310976, "loss": 1.4697, "step": 6089 }, { "epoch": 0.6, "learning_rate": 0.0007271500482592395, "loss": 1.4229, "step": 6090 }, { "epoch": 0.6, "learning_rate": 0.0007268428775051123, "loss": 1.4668, "step": 6091 }, { "epoch": 0.6, "learning_rate": 0.000726535734600033, "loss": 1.4541, "step": 6092 }, { "epoch": 0.6, "learning_rate": 0.0007262286195753153, "loss": 1.519, "step": 6093 }, { "epoch": 0.6, "learning_rate": 0.0007259215324622705, "loss": 1.5518, "step": 6094 }, { "epoch": 0.6, "learning_rate": 0.0007256144732922067, "loss": 1.583, "step": 6095 }, { "epoch": 0.6, "learning_rate": 0.0007253074420964298, "loss": 1.4365, "step": 6096 }, { "epoch": 0.6, "learning_rate": 0.0007250004389062419, "loss": 1.4688, "step": 6097 }, { "epoch": 0.6, "learning_rate": 0.0007246934637529427, "loss": 1.4053, "step": 6098 }, { "epoch": 0.6, "learning_rate": 0.0007243865166678297, "loss": 1.6045, "step": 6099 }, { "epoch": 0.6, "learning_rate": 0.0007240795976821964, "loss": 1.5439, "step": 6100 }, { "epoch": 0.6, "learning_rate": 0.0007237727068273341, "loss": 1.5225, "step": 6101 }, { "epoch": 0.6, "learning_rate": 0.0007234658441345311, "loss": 1.5146, "step": 6102 }, { "epoch": 0.6, "learning_rate": 0.0007231590096350727, "loss": 1.4756, "step": 6103 }, { "epoch": 0.6, "learning_rate": 0.0007228522033602416, "loss": 1.397, "step": 6104 }, { "epoch": 0.6, "learning_rate": 0.0007225454253413175, "loss": 1.4346, "step": 6105 }, { "epoch": 0.6, "learning_rate": 0.0007222386756095769, "loss": 1.4854, "step": 6106 }, { "epoch": 0.6, "learning_rate": 0.000721931954196294, "loss": 1.4902, "step": 6107 }, { "epoch": 0.6, "learning_rate": 0.0007216252611327394, "loss": 1.4834, "step": 6108 }, { "epoch": 0.6, "learning_rate": 0.0007213185964501818, "loss": 1.3613, "step": 6109 }, { "epoch": 0.6, "learning_rate": 0.0007210119601798859, "loss": 1.3916, "step": 6110 }, { "epoch": 0.6, "learning_rate": 0.0007207053523531144, "loss": 1.4766, "step": 6111 }, { "epoch": 0.6, "learning_rate": 0.0007203987730011267, "loss": 1.4746, "step": 6112 }, { "epoch": 0.6, "learning_rate": 0.0007200922221551791, "loss": 1.5537, "step": 6113 }, { "epoch": 0.6, "learning_rate": 0.0007197856998465253, "loss": 1.4443, "step": 6114 }, { "epoch": 0.6, "learning_rate": 0.0007194792061064162, "loss": 1.4609, "step": 6115 }, { "epoch": 0.6, "learning_rate": 0.0007191727409660993, "loss": 1.4004, "step": 6116 }, { "epoch": 0.6, "learning_rate": 0.0007188663044568194, "loss": 1.5586, "step": 6117 }, { "epoch": 0.6, "learning_rate": 0.0007185598966098191, "loss": 1.4033, "step": 6118 }, { "epoch": 0.6, "learning_rate": 0.0007182535174563368, "loss": 1.4873, "step": 6119 }, { "epoch": 0.6, "learning_rate": 0.000717947167027609, "loss": 1.4932, "step": 6120 }, { "epoch": 0.6, "learning_rate": 0.0007176408453548686, "loss": 1.6367, "step": 6121 }, { "epoch": 0.6, "learning_rate": 0.0007173345524693461, "loss": 1.5635, "step": 6122 }, { "epoch": 0.6, "learning_rate": 0.0007170282884022687, "loss": 1.3984, "step": 6123 }, { "epoch": 0.6, "learning_rate": 0.0007167220531848609, "loss": 1.4443, "step": 6124 }, { "epoch": 0.6, "learning_rate": 0.000716415846848344, "loss": 1.583, "step": 6125 }, { "epoch": 0.6, "learning_rate": 0.0007161096694239366, "loss": 1.6084, "step": 6126 }, { "epoch": 0.6, "learning_rate": 0.000715803520942854, "loss": 1.4668, "step": 6127 }, { "epoch": 0.6, "learning_rate": 0.000715497401436309, "loss": 1.458, "step": 6128 }, { "epoch": 0.6, "learning_rate": 0.0007151913109355114, "loss": 1.4971, "step": 6129 }, { "epoch": 0.6, "learning_rate": 0.0007148852494716676, "loss": 1.4775, "step": 6130 }, { "epoch": 0.6, "learning_rate": 0.0007145792170759811, "loss": 1.5693, "step": 6131 }, { "epoch": 0.6, "learning_rate": 0.0007142732137796535, "loss": 1.4443, "step": 6132 }, { "epoch": 0.6, "learning_rate": 0.0007139672396138818, "loss": 1.5225, "step": 6133 }, { "epoch": 0.6, "learning_rate": 0.0007136612946098614, "loss": 1.4424, "step": 6134 }, { "epoch": 0.6, "learning_rate": 0.0007133553787987838, "loss": 1.459, "step": 6135 }, { "epoch": 0.6, "learning_rate": 0.000713049492211838, "loss": 1.584, "step": 6136 }, { "epoch": 0.6, "learning_rate": 0.0007127436348802098, "loss": 1.4492, "step": 6137 }, { "epoch": 0.61, "learning_rate": 0.0007124378068350822, "loss": 1.4414, "step": 6138 }, { "epoch": 0.61, "learning_rate": 0.000712132008107635, "loss": 1.498, "step": 6139 }, { "epoch": 0.61, "learning_rate": 0.0007118262387290453, "loss": 1.5029, "step": 6140 }, { "epoch": 0.61, "learning_rate": 0.0007115204987304869, "loss": 1.3789, "step": 6141 }, { "epoch": 0.61, "learning_rate": 0.000711214788143131, "loss": 1.6045, "step": 6142 }, { "epoch": 0.61, "learning_rate": 0.0007109091069981455, "loss": 1.457, "step": 6143 }, { "epoch": 0.61, "learning_rate": 0.0007106034553266952, "loss": 1.4092, "step": 6144 }, { "epoch": 0.61, "learning_rate": 0.000710297833159942, "loss": 1.4932, "step": 6145 }, { "epoch": 0.61, "learning_rate": 0.000709992240529045, "loss": 1.4414, "step": 6146 }, { "epoch": 0.61, "learning_rate": 0.00070968667746516, "loss": 1.4424, "step": 6147 }, { "epoch": 0.61, "learning_rate": 0.0007093811439994402, "loss": 1.5059, "step": 6148 }, { "epoch": 0.61, "learning_rate": 0.000709075640163035, "loss": 1.5195, "step": 6149 }, { "epoch": 0.61, "learning_rate": 0.0007087701659870915, "loss": 1.5244, "step": 6150 }, { "epoch": 0.61, "learning_rate": 0.0007084647215027538, "loss": 1.4971, "step": 6151 }, { "epoch": 0.61, "learning_rate": 0.0007081593067411627, "loss": 1.4854, "step": 6152 }, { "epoch": 0.61, "learning_rate": 0.0007078539217334557, "loss": 1.3877, "step": 6153 }, { "epoch": 0.61, "learning_rate": 0.0007075485665107676, "loss": 1.4717, "step": 6154 }, { "epoch": 0.61, "learning_rate": 0.0007072432411042302, "loss": 1.54, "step": 6155 }, { "epoch": 0.61, "learning_rate": 0.0007069379455449723, "loss": 1.5312, "step": 6156 }, { "epoch": 0.61, "learning_rate": 0.0007066326798641195, "loss": 1.498, "step": 6157 }, { "epoch": 0.61, "learning_rate": 0.0007063274440927941, "loss": 1.5508, "step": 6158 }, { "epoch": 0.61, "learning_rate": 0.0007060222382621159, "loss": 1.4199, "step": 6159 }, { "epoch": 0.61, "learning_rate": 0.0007057170624032017, "loss": 1.332, "step": 6160 }, { "epoch": 0.61, "learning_rate": 0.0007054119165471643, "loss": 1.4795, "step": 6161 }, { "epoch": 0.61, "learning_rate": 0.0007051068007251144, "loss": 1.4717, "step": 6162 }, { "epoch": 0.61, "learning_rate": 0.0007048017149681596, "loss": 1.4619, "step": 6163 }, { "epoch": 0.61, "learning_rate": 0.0007044966593074036, "loss": 1.4795, "step": 6164 }, { "epoch": 0.61, "learning_rate": 0.000704191633773948, "loss": 1.4971, "step": 6165 }, { "epoch": 0.61, "learning_rate": 0.0007038866383988906, "loss": 1.5195, "step": 6166 }, { "epoch": 0.61, "learning_rate": 0.0007035816732133264, "loss": 1.5732, "step": 6167 }, { "epoch": 0.61, "learning_rate": 0.0007032767382483478, "loss": 1.4697, "step": 6168 }, { "epoch": 0.61, "learning_rate": 0.0007029718335350434, "loss": 1.5068, "step": 6169 }, { "epoch": 0.61, "learning_rate": 0.0007026669591044992, "loss": 1.5215, "step": 6170 }, { "epoch": 0.61, "learning_rate": 0.0007023621149877975, "loss": 1.4971, "step": 6171 }, { "epoch": 0.61, "learning_rate": 0.0007020573012160183, "loss": 1.5273, "step": 6172 }, { "epoch": 0.61, "learning_rate": 0.0007017525178202379, "loss": 1.4971, "step": 6173 }, { "epoch": 0.61, "learning_rate": 0.0007014477648315297, "loss": 1.4619, "step": 6174 }, { "epoch": 0.61, "learning_rate": 0.0007011430422809642, "loss": 1.4893, "step": 6175 }, { "epoch": 0.61, "learning_rate": 0.0007008383501996083, "loss": 1.5186, "step": 6176 }, { "epoch": 0.61, "learning_rate": 0.0007005336886185265, "loss": 1.5684, "step": 6177 }, { "epoch": 0.61, "learning_rate": 0.0007002290575687797, "loss": 1.4785, "step": 6178 }, { "epoch": 0.61, "learning_rate": 0.0006999244570814259, "loss": 1.6338, "step": 6179 }, { "epoch": 0.61, "learning_rate": 0.0006996198871875195, "loss": 1.5684, "step": 6180 }, { "epoch": 0.61, "learning_rate": 0.0006993153479181127, "loss": 1.3564, "step": 6181 }, { "epoch": 0.61, "learning_rate": 0.0006990108393042536, "loss": 1.5127, "step": 6182 }, { "epoch": 0.61, "learning_rate": 0.0006987063613769875, "loss": 1.4746, "step": 6183 }, { "epoch": 0.61, "learning_rate": 0.000698401914167357, "loss": 1.5264, "step": 6184 }, { "epoch": 0.61, "learning_rate": 0.0006980974977064012, "loss": 1.5605, "step": 6185 }, { "epoch": 0.61, "learning_rate": 0.0006977931120251561, "loss": 1.498, "step": 6186 }, { "epoch": 0.61, "learning_rate": 0.0006974887571546545, "loss": 1.3682, "step": 6187 }, { "epoch": 0.61, "learning_rate": 0.0006971844331259263, "loss": 1.2988, "step": 6188 }, { "epoch": 0.61, "learning_rate": 0.0006968801399699978, "loss": 1.6582, "step": 6189 }, { "epoch": 0.61, "learning_rate": 0.0006965758777178926, "loss": 1.4229, "step": 6190 }, { "epoch": 0.61, "learning_rate": 0.0006962716464006311, "loss": 1.4971, "step": 6191 }, { "epoch": 0.61, "learning_rate": 0.0006959674460492303, "loss": 1.4424, "step": 6192 }, { "epoch": 0.61, "learning_rate": 0.0006956632766947044, "loss": 1.4404, "step": 6193 }, { "epoch": 0.61, "learning_rate": 0.000695359138368064, "loss": 1.5537, "step": 6194 }, { "epoch": 0.61, "learning_rate": 0.0006950550311003167, "loss": 1.4893, "step": 6195 }, { "epoch": 0.61, "learning_rate": 0.0006947509549224673, "loss": 1.4512, "step": 6196 }, { "epoch": 0.61, "learning_rate": 0.0006944469098655168, "loss": 1.4385, "step": 6197 }, { "epoch": 0.61, "learning_rate": 0.0006941428959604634, "loss": 1.397, "step": 6198 }, { "epoch": 0.61, "learning_rate": 0.0006938389132383022, "loss": 1.4014, "step": 6199 }, { "epoch": 0.61, "learning_rate": 0.000693534961730025, "loss": 1.5469, "step": 6200 }, { "epoch": 0.61, "learning_rate": 0.0006932310414666203, "loss": 1.3828, "step": 6201 }, { "epoch": 0.61, "learning_rate": 0.0006929271524790737, "loss": 1.5059, "step": 6202 }, { "epoch": 0.61, "learning_rate": 0.0006926232947983673, "loss": 1.4805, "step": 6203 }, { "epoch": 0.61, "learning_rate": 0.0006923194684554802, "loss": 1.4385, "step": 6204 }, { "epoch": 0.61, "learning_rate": 0.0006920156734813882, "loss": 1.457, "step": 6205 }, { "epoch": 0.61, "learning_rate": 0.0006917119099070639, "loss": 1.4004, "step": 6206 }, { "epoch": 0.61, "learning_rate": 0.0006914081777634769, "loss": 1.4756, "step": 6207 }, { "epoch": 0.61, "learning_rate": 0.0006911044770815932, "loss": 1.5137, "step": 6208 }, { "epoch": 0.61, "learning_rate": 0.0006908008078923762, "loss": 1.4863, "step": 6209 }, { "epoch": 0.61, "learning_rate": 0.0006904971702267855, "loss": 1.5342, "step": 6210 }, { "epoch": 0.61, "learning_rate": 0.0006901935641157777, "loss": 1.5215, "step": 6211 }, { "epoch": 0.61, "learning_rate": 0.0006898899895903064, "loss": 1.5547, "step": 6212 }, { "epoch": 0.61, "learning_rate": 0.0006895864466813216, "loss": 1.5195, "step": 6213 }, { "epoch": 0.61, "learning_rate": 0.0006892829354197702, "loss": 1.625, "step": 6214 }, { "epoch": 0.61, "learning_rate": 0.000688979455836596, "loss": 1.5732, "step": 6215 }, { "epoch": 0.61, "learning_rate": 0.0006886760079627397, "loss": 1.6553, "step": 6216 }, { "epoch": 0.61, "learning_rate": 0.0006883725918291379, "loss": 1.5576, "step": 6217 }, { "epoch": 0.61, "learning_rate": 0.0006880692074667257, "loss": 1.542, "step": 6218 }, { "epoch": 0.61, "learning_rate": 0.000687765854906433, "loss": 1.4385, "step": 6219 }, { "epoch": 0.61, "learning_rate": 0.0006874625341791875, "loss": 1.4297, "step": 6220 }, { "epoch": 0.61, "learning_rate": 0.0006871592453159139, "loss": 1.4805, "step": 6221 }, { "epoch": 0.61, "learning_rate": 0.0006868559883475329, "loss": 1.5811, "step": 6222 }, { "epoch": 0.61, "learning_rate": 0.0006865527633049626, "loss": 1.4268, "step": 6223 }, { "epoch": 0.61, "learning_rate": 0.0006862495702191171, "loss": 1.5576, "step": 6224 }, { "epoch": 0.61, "learning_rate": 0.0006859464091209082, "loss": 1.5625, "step": 6225 }, { "epoch": 0.61, "learning_rate": 0.0006856432800412432, "loss": 1.4785, "step": 6226 }, { "epoch": 0.61, "learning_rate": 0.0006853401830110277, "loss": 1.4238, "step": 6227 }, { "epoch": 0.61, "learning_rate": 0.0006850371180611627, "loss": 1.4912, "step": 6228 }, { "epoch": 0.61, "learning_rate": 0.0006847340852225466, "loss": 1.4521, "step": 6229 }, { "epoch": 0.61, "learning_rate": 0.0006844310845260743, "loss": 1.5176, "step": 6230 }, { "epoch": 0.61, "learning_rate": 0.0006841281160026375, "loss": 1.4932, "step": 6231 }, { "epoch": 0.61, "learning_rate": 0.0006838251796831245, "loss": 1.4834, "step": 6232 }, { "epoch": 0.61, "learning_rate": 0.0006835222755984205, "loss": 1.4336, "step": 6233 }, { "epoch": 0.61, "learning_rate": 0.0006832194037794074, "loss": 1.459, "step": 6234 }, { "epoch": 0.61, "learning_rate": 0.0006829165642569634, "loss": 1.4565, "step": 6235 }, { "epoch": 0.61, "learning_rate": 0.0006826137570619641, "loss": 1.5566, "step": 6236 }, { "epoch": 0.61, "learning_rate": 0.0006823109822252818, "loss": 1.5176, "step": 6237 }, { "epoch": 0.61, "learning_rate": 0.0006820082397777842, "loss": 1.5928, "step": 6238 }, { "epoch": 0.62, "learning_rate": 0.0006817055297503372, "loss": 1.5781, "step": 6239 }, { "epoch": 0.62, "learning_rate": 0.0006814028521738026, "loss": 1.5166, "step": 6240 }, { "epoch": 0.62, "learning_rate": 0.0006811002070790395, "loss": 1.5576, "step": 6241 }, { "epoch": 0.62, "learning_rate": 0.0006807975944969029, "loss": 1.5615, "step": 6242 }, { "epoch": 0.62, "learning_rate": 0.0006804950144582454, "loss": 1.5049, "step": 6243 }, { "epoch": 0.62, "learning_rate": 0.0006801924669939154, "loss": 1.6699, "step": 6244 }, { "epoch": 0.62, "learning_rate": 0.0006798899521347583, "loss": 1.5986, "step": 6245 }, { "epoch": 0.62, "learning_rate": 0.0006795874699116167, "loss": 1.4092, "step": 6246 }, { "epoch": 0.62, "learning_rate": 0.0006792850203553289, "loss": 1.3887, "step": 6247 }, { "epoch": 0.62, "learning_rate": 0.0006789826034967307, "loss": 1.457, "step": 6248 }, { "epoch": 0.62, "learning_rate": 0.0006786802193666539, "loss": 1.4824, "step": 6249 }, { "epoch": 0.62, "learning_rate": 0.0006783778679959278, "loss": 1.4375, "step": 6250 }, { "epoch": 0.62, "learning_rate": 0.0006780755494153778, "loss": 1.4502, "step": 6251 }, { "epoch": 0.62, "learning_rate": 0.0006777732636558257, "loss": 1.4365, "step": 6252 }, { "epoch": 0.62, "learning_rate": 0.0006774710107480906, "loss": 1.3872, "step": 6253 }, { "epoch": 0.62, "learning_rate": 0.0006771687907229876, "loss": 1.4492, "step": 6254 }, { "epoch": 0.62, "learning_rate": 0.0006768666036113291, "loss": 1.5381, "step": 6255 }, { "epoch": 0.62, "learning_rate": 0.0006765644494439239, "loss": 1.4717, "step": 6256 }, { "epoch": 0.62, "learning_rate": 0.0006762623282515768, "loss": 1.4893, "step": 6257 }, { "epoch": 0.62, "learning_rate": 0.0006759602400650902, "loss": 1.5391, "step": 6258 }, { "epoch": 0.62, "learning_rate": 0.000675658184915263, "loss": 1.4854, "step": 6259 }, { "epoch": 0.62, "learning_rate": 0.0006753561628328901, "loss": 1.4805, "step": 6260 }, { "epoch": 0.62, "learning_rate": 0.0006750541738487634, "loss": 1.3916, "step": 6261 }, { "epoch": 0.62, "learning_rate": 0.0006747522179936717, "loss": 1.502, "step": 6262 }, { "epoch": 0.62, "learning_rate": 0.0006744502952983999, "loss": 1.5312, "step": 6263 }, { "epoch": 0.62, "learning_rate": 0.00067414840579373, "loss": 1.5098, "step": 6264 }, { "epoch": 0.62, "learning_rate": 0.00067384654951044, "loss": 1.542, "step": 6265 }, { "epoch": 0.62, "learning_rate": 0.0006735447264793053, "loss": 1.4287, "step": 6266 }, { "epoch": 0.62, "learning_rate": 0.0006732429367310971, "loss": 1.5488, "step": 6267 }, { "epoch": 0.62, "learning_rate": 0.000672941180296584, "loss": 1.5771, "step": 6268 }, { "epoch": 0.62, "learning_rate": 0.0006726394572065307, "loss": 1.3779, "step": 6269 }, { "epoch": 0.62, "learning_rate": 0.0006723377674916986, "loss": 1.4619, "step": 6270 }, { "epoch": 0.62, "learning_rate": 0.0006720361111828458, "loss": 1.5586, "step": 6271 }, { "epoch": 0.62, "learning_rate": 0.0006717344883107268, "loss": 1.3965, "step": 6272 }, { "epoch": 0.62, "learning_rate": 0.0006714328989060927, "loss": 1.4658, "step": 6273 }, { "epoch": 0.62, "learning_rate": 0.0006711313429996915, "loss": 1.5264, "step": 6274 }, { "epoch": 0.62, "learning_rate": 0.0006708298206222675, "loss": 1.4805, "step": 6275 }, { "epoch": 0.62, "learning_rate": 0.0006705283318045615, "loss": 1.4072, "step": 6276 }, { "epoch": 0.62, "learning_rate": 0.0006702268765773112, "loss": 1.4766, "step": 6277 }, { "epoch": 0.62, "learning_rate": 0.0006699254549712508, "loss": 1.4121, "step": 6278 }, { "epoch": 0.62, "learning_rate": 0.0006696240670171109, "loss": 1.4424, "step": 6279 }, { "epoch": 0.62, "learning_rate": 0.0006693227127456187, "loss": 1.4365, "step": 6280 }, { "epoch": 0.62, "learning_rate": 0.000669021392187498, "loss": 1.6396, "step": 6281 }, { "epoch": 0.62, "learning_rate": 0.0006687201053734693, "loss": 1.4746, "step": 6282 }, { "epoch": 0.62, "learning_rate": 0.0006684188523342493, "loss": 1.623, "step": 6283 }, { "epoch": 0.62, "learning_rate": 0.0006681176331005519, "loss": 1.5264, "step": 6284 }, { "epoch": 0.62, "learning_rate": 0.0006678164477030865, "loss": 1.4189, "step": 6285 }, { "epoch": 0.62, "learning_rate": 0.0006675152961725604, "loss": 1.4756, "step": 6286 }, { "epoch": 0.62, "learning_rate": 0.0006672141785396764, "loss": 1.3403, "step": 6287 }, { "epoch": 0.62, "learning_rate": 0.0006669130948351343, "loss": 1.4111, "step": 6288 }, { "epoch": 0.62, "learning_rate": 0.0006666120450896303, "loss": 1.4287, "step": 6289 }, { "epoch": 0.62, "learning_rate": 0.000666311029333857, "loss": 1.458, "step": 6290 }, { "epoch": 0.62, "learning_rate": 0.000666010047598504, "loss": 1.4473, "step": 6291 }, { "epoch": 0.62, "learning_rate": 0.0006657090999142569, "loss": 1.3945, "step": 6292 }, { "epoch": 0.62, "learning_rate": 0.0006654081863117985, "loss": 1.5146, "step": 6293 }, { "epoch": 0.62, "learning_rate": 0.000665107306821807, "loss": 1.5254, "step": 6294 }, { "epoch": 0.62, "learning_rate": 0.0006648064614749581, "loss": 1.5137, "step": 6295 }, { "epoch": 0.62, "learning_rate": 0.0006645056503019238, "loss": 1.3818, "step": 6296 }, { "epoch": 0.62, "learning_rate": 0.0006642048733333727, "loss": 1.4688, "step": 6297 }, { "epoch": 0.62, "learning_rate": 0.0006639041305999693, "loss": 1.5361, "step": 6298 }, { "epoch": 0.62, "learning_rate": 0.0006636034221323754, "loss": 1.3613, "step": 6299 }, { "epoch": 0.62, "learning_rate": 0.0006633027479612491, "loss": 1.5361, "step": 6300 }, { "epoch": 0.62, "learning_rate": 0.0006630021081172449, "loss": 1.3926, "step": 6301 }, { "epoch": 0.62, "learning_rate": 0.0006627015026310132, "loss": 1.4414, "step": 6302 }, { "epoch": 0.62, "learning_rate": 0.000662400931533202, "loss": 1.4531, "step": 6303 }, { "epoch": 0.62, "learning_rate": 0.0006621003948544553, "loss": 1.5371, "step": 6304 }, { "epoch": 0.62, "learning_rate": 0.0006617998926254133, "loss": 1.4463, "step": 6305 }, { "epoch": 0.62, "learning_rate": 0.000661499424876713, "loss": 1.5654, "step": 6306 }, { "epoch": 0.62, "learning_rate": 0.000661198991638988, "loss": 1.4492, "step": 6307 }, { "epoch": 0.62, "learning_rate": 0.0006608985929428679, "loss": 1.5566, "step": 6308 }, { "epoch": 0.62, "learning_rate": 0.0006605982288189794, "loss": 1.4775, "step": 6309 }, { "epoch": 0.62, "learning_rate": 0.0006602978992979454, "loss": 1.4004, "step": 6310 }, { "epoch": 0.62, "learning_rate": 0.0006599976044103851, "loss": 1.5649, "step": 6311 }, { "epoch": 0.62, "learning_rate": 0.0006596973441869141, "loss": 1.5117, "step": 6312 }, { "epoch": 0.62, "learning_rate": 0.0006593971186581451, "loss": 1.4375, "step": 6313 }, { "epoch": 0.62, "learning_rate": 0.0006590969278546866, "loss": 1.4531, "step": 6314 }, { "epoch": 0.62, "learning_rate": 0.0006587967718071437, "loss": 1.4746, "step": 6315 }, { "epoch": 0.62, "learning_rate": 0.0006584966505461183, "loss": 1.5576, "step": 6316 }, { "epoch": 0.62, "learning_rate": 0.0006581965641022084, "loss": 1.5088, "step": 6317 }, { "epoch": 0.62, "learning_rate": 0.0006578965125060084, "loss": 1.4551, "step": 6318 }, { "epoch": 0.62, "learning_rate": 0.0006575964957881096, "loss": 1.4907, "step": 6319 }, { "epoch": 0.62, "learning_rate": 0.0006572965139790992, "loss": 1.4648, "step": 6320 }, { "epoch": 0.62, "learning_rate": 0.0006569965671095613, "loss": 1.5156, "step": 6321 }, { "epoch": 0.62, "learning_rate": 0.0006566966552100762, "loss": 1.4004, "step": 6322 }, { "epoch": 0.62, "learning_rate": 0.0006563967783112205, "loss": 1.5811, "step": 6323 }, { "epoch": 0.62, "learning_rate": 0.0006560969364435677, "loss": 1.3906, "step": 6324 }, { "epoch": 0.62, "learning_rate": 0.0006557971296376869, "loss": 1.5469, "step": 6325 }, { "epoch": 0.62, "learning_rate": 0.0006554973579241445, "loss": 1.4834, "step": 6326 }, { "epoch": 0.62, "learning_rate": 0.0006551976213335029, "loss": 1.584, "step": 6327 }, { "epoch": 0.62, "learning_rate": 0.0006548979198963212, "loss": 1.4072, "step": 6328 }, { "epoch": 0.62, "learning_rate": 0.0006545982536431545, "loss": 1.627, "step": 6329 }, { "epoch": 0.62, "learning_rate": 0.0006542986226045546, "loss": 1.5391, "step": 6330 }, { "epoch": 0.62, "learning_rate": 0.0006539990268110695, "loss": 1.5029, "step": 6331 }, { "epoch": 0.62, "learning_rate": 0.0006536994662932438, "loss": 1.4072, "step": 6332 }, { "epoch": 0.62, "learning_rate": 0.0006533999410816185, "loss": 1.4287, "step": 6333 }, { "epoch": 0.62, "learning_rate": 0.0006531004512067309, "loss": 1.4971, "step": 6334 }, { "epoch": 0.62, "learning_rate": 0.0006528009966991149, "loss": 1.4395, "step": 6335 }, { "epoch": 0.62, "learning_rate": 0.0006525015775893, "loss": 1.416, "step": 6336 }, { "epoch": 0.62, "learning_rate": 0.0006522021939078136, "loss": 1.6855, "step": 6337 }, { "epoch": 0.62, "learning_rate": 0.000651902845685178, "loss": 1.4619, "step": 6338 }, { "epoch": 0.62, "learning_rate": 0.0006516035329519128, "loss": 1.3828, "step": 6339 }, { "epoch": 0.62, "learning_rate": 0.0006513042557385336, "loss": 1.502, "step": 6340 }, { "epoch": 0.63, "learning_rate": 0.0006510050140755523, "loss": 1.5029, "step": 6341 }, { "epoch": 0.63, "learning_rate": 0.0006507058079934776, "loss": 1.4043, "step": 6342 }, { "epoch": 0.63, "learning_rate": 0.000650406637522814, "loss": 1.5283, "step": 6343 }, { "epoch": 0.63, "learning_rate": 0.000650107502694063, "loss": 1.4043, "step": 6344 }, { "epoch": 0.63, "learning_rate": 0.0006498084035377217, "loss": 1.3232, "step": 6345 }, { "epoch": 0.63, "learning_rate": 0.0006495093400842845, "loss": 1.6689, "step": 6346 }, { "epoch": 0.63, "learning_rate": 0.0006492103123642415, "loss": 1.4766, "step": 6347 }, { "epoch": 0.63, "learning_rate": 0.0006489113204080791, "loss": 1.5762, "step": 6348 }, { "epoch": 0.63, "learning_rate": 0.0006486123642462809, "loss": 1.5234, "step": 6349 }, { "epoch": 0.63, "learning_rate": 0.0006483134439093252, "loss": 1.3613, "step": 6350 }, { "epoch": 0.63, "learning_rate": 0.0006480145594276885, "loss": 1.5381, "step": 6351 }, { "epoch": 0.63, "learning_rate": 0.0006477157108318425, "loss": 1.501, "step": 6352 }, { "epoch": 0.63, "learning_rate": 0.0006474168981522554, "loss": 1.582, "step": 6353 }, { "epoch": 0.63, "learning_rate": 0.0006471181214193925, "loss": 1.4668, "step": 6354 }, { "epoch": 0.63, "learning_rate": 0.0006468193806637141, "loss": 1.4404, "step": 6355 }, { "epoch": 0.63, "learning_rate": 0.0006465206759156778, "loss": 1.5088, "step": 6356 }, { "epoch": 0.63, "learning_rate": 0.0006462220072057376, "loss": 1.4023, "step": 6357 }, { "epoch": 0.63, "learning_rate": 0.0006459233745643434, "loss": 1.4629, "step": 6358 }, { "epoch": 0.63, "learning_rate": 0.0006456247780219411, "loss": 1.4854, "step": 6359 }, { "epoch": 0.63, "learning_rate": 0.0006453262176089738, "loss": 1.4082, "step": 6360 }, { "epoch": 0.63, "learning_rate": 0.0006450276933558805, "loss": 1.5342, "step": 6361 }, { "epoch": 0.63, "learning_rate": 0.0006447292052930965, "loss": 1.4824, "step": 6362 }, { "epoch": 0.63, "learning_rate": 0.0006444307534510529, "loss": 1.4346, "step": 6363 }, { "epoch": 0.63, "learning_rate": 0.0006441323378601783, "loss": 1.4756, "step": 6364 }, { "epoch": 0.63, "learning_rate": 0.0006438339585508964, "loss": 1.6279, "step": 6365 }, { "epoch": 0.63, "learning_rate": 0.000643535615553628, "loss": 1.4189, "step": 6366 }, { "epoch": 0.63, "learning_rate": 0.0006432373088987897, "loss": 1.4541, "step": 6367 }, { "epoch": 0.63, "learning_rate": 0.0006429390386167947, "loss": 1.5273, "step": 6368 }, { "epoch": 0.63, "learning_rate": 0.0006426408047380525, "loss": 1.498, "step": 6369 }, { "epoch": 0.63, "learning_rate": 0.0006423426072929688, "loss": 1.5537, "step": 6370 }, { "epoch": 0.63, "learning_rate": 0.0006420444463119452, "loss": 1.4473, "step": 6371 }, { "epoch": 0.63, "learning_rate": 0.0006417463218253805, "loss": 1.3179, "step": 6372 }, { "epoch": 0.63, "learning_rate": 0.0006414482338636687, "loss": 1.5762, "step": 6373 }, { "epoch": 0.63, "learning_rate": 0.000641150182457201, "loss": 1.3916, "step": 6374 }, { "epoch": 0.63, "learning_rate": 0.0006408521676363644, "loss": 1.3325, "step": 6375 }, { "epoch": 0.63, "learning_rate": 0.0006405541894315421, "loss": 1.4033, "step": 6376 }, { "epoch": 0.63, "learning_rate": 0.0006402562478731137, "loss": 1.5518, "step": 6377 }, { "epoch": 0.63, "learning_rate": 0.0006399583429914553, "loss": 1.3955, "step": 6378 }, { "epoch": 0.63, "learning_rate": 0.000639660474816939, "loss": 1.3799, "step": 6379 }, { "epoch": 0.63, "learning_rate": 0.0006393626433799331, "loss": 1.3535, "step": 6380 }, { "epoch": 0.63, "learning_rate": 0.0006390648487108024, "loss": 1.335, "step": 6381 }, { "epoch": 0.63, "learning_rate": 0.0006387670908399076, "loss": 1.4707, "step": 6382 }, { "epoch": 0.63, "learning_rate": 0.0006384693697976061, "loss": 1.5176, "step": 6383 }, { "epoch": 0.63, "learning_rate": 0.000638171685614251, "loss": 1.5322, "step": 6384 }, { "epoch": 0.63, "learning_rate": 0.000637874038320192, "loss": 1.4697, "step": 6385 }, { "epoch": 0.63, "learning_rate": 0.0006375764279457751, "loss": 1.5684, "step": 6386 }, { "epoch": 0.63, "learning_rate": 0.0006372788545213425, "loss": 1.458, "step": 6387 }, { "epoch": 0.63, "learning_rate": 0.0006369813180772325, "loss": 1.7432, "step": 6388 }, { "epoch": 0.63, "learning_rate": 0.0006366838186437796, "loss": 1.3994, "step": 6389 }, { "epoch": 0.63, "learning_rate": 0.0006363863562513145, "loss": 1.4688, "step": 6390 }, { "epoch": 0.63, "learning_rate": 0.0006360889309301646, "loss": 1.458, "step": 6391 }, { "epoch": 0.63, "learning_rate": 0.0006357915427106529, "loss": 1.5508, "step": 6392 }, { "epoch": 0.63, "learning_rate": 0.0006354941916230987, "loss": 1.5244, "step": 6393 }, { "epoch": 0.63, "learning_rate": 0.000635196877697818, "loss": 1.5322, "step": 6394 }, { "epoch": 0.63, "learning_rate": 0.0006348996009651225, "loss": 1.5752, "step": 6395 }, { "epoch": 0.63, "learning_rate": 0.0006346023614553203, "loss": 1.6006, "step": 6396 }, { "epoch": 0.63, "learning_rate": 0.0006343051591987159, "loss": 1.4561, "step": 6397 }, { "epoch": 0.63, "learning_rate": 0.0006340079942256096, "loss": 1.458, "step": 6398 }, { "epoch": 0.63, "learning_rate": 0.0006337108665662983, "loss": 1.3623, "step": 6399 }, { "epoch": 0.63, "learning_rate": 0.0006334137762510748, "loss": 1.4512, "step": 6400 }, { "epoch": 0.63, "learning_rate": 0.0006331167233102282, "loss": 1.3916, "step": 6401 }, { "epoch": 0.63, "learning_rate": 0.0006328197077740437, "loss": 1.5371, "step": 6402 }, { "epoch": 0.63, "learning_rate": 0.0006325227296728028, "loss": 1.4062, "step": 6403 }, { "epoch": 0.63, "learning_rate": 0.0006322257890367831, "loss": 1.4082, "step": 6404 }, { "epoch": 0.63, "learning_rate": 0.0006319288858962586, "loss": 1.3867, "step": 6405 }, { "epoch": 0.63, "learning_rate": 0.0006316320202814998, "loss": 1.4043, "step": 6406 }, { "epoch": 0.63, "learning_rate": 0.0006313351922227719, "loss": 1.5674, "step": 6407 }, { "epoch": 0.63, "learning_rate": 0.0006310384017503376, "loss": 1.5303, "step": 6408 }, { "epoch": 0.63, "learning_rate": 0.0006307416488944555, "loss": 1.2109, "step": 6409 }, { "epoch": 0.63, "learning_rate": 0.0006304449336853805, "loss": 1.5176, "step": 6410 }, { "epoch": 0.63, "learning_rate": 0.0006301482561533631, "loss": 1.5, "step": 6411 }, { "epoch": 0.63, "learning_rate": 0.0006298516163286506, "loss": 1.3789, "step": 6412 }, { "epoch": 0.63, "learning_rate": 0.0006295550142414861, "loss": 1.459, "step": 6413 }, { "epoch": 0.63, "learning_rate": 0.0006292584499221087, "loss": 1.3799, "step": 6414 }, { "epoch": 0.63, "learning_rate": 0.0006289619234007542, "loss": 1.5635, "step": 6415 }, { "epoch": 0.63, "learning_rate": 0.000628665434707654, "loss": 1.4033, "step": 6416 }, { "epoch": 0.63, "learning_rate": 0.000628368983873036, "loss": 1.5225, "step": 6417 }, { "epoch": 0.63, "learning_rate": 0.0006280725709271236, "loss": 1.5332, "step": 6418 }, { "epoch": 0.63, "learning_rate": 0.0006277761959001376, "loss": 1.5771, "step": 6419 }, { "epoch": 0.63, "learning_rate": 0.0006274798588222939, "loss": 1.3745, "step": 6420 }, { "epoch": 0.63, "learning_rate": 0.0006271835597238045, "loss": 1.542, "step": 6421 }, { "epoch": 0.63, "learning_rate": 0.0006268872986348782, "loss": 1.5137, "step": 6422 }, { "epoch": 0.63, "learning_rate": 0.0006265910755857193, "loss": 1.4922, "step": 6423 }, { "epoch": 0.63, "learning_rate": 0.0006262948906065287, "loss": 1.4629, "step": 6424 }, { "epoch": 0.63, "learning_rate": 0.0006259987437275032, "loss": 1.5225, "step": 6425 }, { "epoch": 0.63, "learning_rate": 0.0006257026349788354, "loss": 1.4834, "step": 6426 }, { "epoch": 0.63, "learning_rate": 0.0006254065643907143, "loss": 1.4443, "step": 6427 }, { "epoch": 0.63, "learning_rate": 0.0006251105319933255, "loss": 1.4883, "step": 6428 }, { "epoch": 0.63, "learning_rate": 0.00062481453781685, "loss": 1.4326, "step": 6429 }, { "epoch": 0.63, "learning_rate": 0.0006245185818914651, "loss": 1.4482, "step": 6430 }, { "epoch": 0.63, "learning_rate": 0.0006242226642473443, "loss": 1.3623, "step": 6431 }, { "epoch": 0.63, "learning_rate": 0.000623926784914657, "loss": 1.4893, "step": 6432 }, { "epoch": 0.63, "learning_rate": 0.000623630943923569, "loss": 1.4238, "step": 6433 }, { "epoch": 0.63, "learning_rate": 0.0006233351413042421, "loss": 1.415, "step": 6434 }, { "epoch": 0.63, "learning_rate": 0.0006230393770868339, "loss": 1.4521, "step": 6435 }, { "epoch": 0.63, "learning_rate": 0.0006227436513014981, "loss": 1.5039, "step": 6436 }, { "epoch": 0.63, "learning_rate": 0.0006224479639783851, "loss": 1.5645, "step": 6437 }, { "epoch": 0.63, "learning_rate": 0.0006221523151476409, "loss": 1.5215, "step": 6438 }, { "epoch": 0.63, "learning_rate": 0.0006218567048394077, "loss": 1.4375, "step": 6439 }, { "epoch": 0.63, "learning_rate": 0.0006215611330838233, "loss": 1.3164, "step": 6440 }, { "epoch": 0.63, "learning_rate": 0.0006212655999110224, "loss": 1.4854, "step": 6441 }, { "epoch": 0.64, "learning_rate": 0.000620970105351135, "loss": 1.6025, "step": 6442 }, { "epoch": 0.64, "learning_rate": 0.0006206746494342879, "loss": 1.4395, "step": 6443 }, { "epoch": 0.64, "learning_rate": 0.0006203792321906032, "loss": 1.5576, "step": 6444 }, { "epoch": 0.64, "learning_rate": 0.0006200838536501994, "loss": 1.5391, "step": 6445 }, { "epoch": 0.64, "learning_rate": 0.0006197885138431913, "loss": 1.4033, "step": 6446 }, { "epoch": 0.64, "learning_rate": 0.0006194932127996897, "loss": 1.4346, "step": 6447 }, { "epoch": 0.64, "learning_rate": 0.0006191979505498009, "loss": 1.4541, "step": 6448 }, { "epoch": 0.64, "learning_rate": 0.0006189027271236279, "loss": 1.459, "step": 6449 }, { "epoch": 0.64, "learning_rate": 0.0006186075425512692, "loss": 1.3252, "step": 6450 }, { "epoch": 0.64, "learning_rate": 0.0006183123968628197, "loss": 1.4072, "step": 6451 }, { "epoch": 0.64, "learning_rate": 0.0006180172900883703, "loss": 1.5273, "step": 6452 }, { "epoch": 0.64, "learning_rate": 0.0006177222222580077, "loss": 1.4717, "step": 6453 }, { "epoch": 0.64, "learning_rate": 0.000617427193401815, "loss": 1.4141, "step": 6454 }, { "epoch": 0.64, "learning_rate": 0.000617132203549871, "loss": 1.3926, "step": 6455 }, { "epoch": 0.64, "learning_rate": 0.0006168372527322508, "loss": 1.4814, "step": 6456 }, { "epoch": 0.64, "learning_rate": 0.0006165423409790254, "loss": 1.4482, "step": 6457 }, { "epoch": 0.64, "learning_rate": 0.0006162474683202615, "loss": 1.3213, "step": 6458 }, { "epoch": 0.64, "learning_rate": 0.0006159526347860223, "loss": 1.4482, "step": 6459 }, { "epoch": 0.64, "learning_rate": 0.0006156578404063667, "loss": 1.5078, "step": 6460 }, { "epoch": 0.64, "learning_rate": 0.0006153630852113499, "loss": 1.4629, "step": 6461 }, { "epoch": 0.64, "learning_rate": 0.000615068369231023, "loss": 1.373, "step": 6462 }, { "epoch": 0.64, "learning_rate": 0.0006147736924954327, "loss": 1.4482, "step": 6463 }, { "epoch": 0.64, "learning_rate": 0.0006144790550346221, "loss": 1.3848, "step": 6464 }, { "epoch": 0.64, "learning_rate": 0.0006141844568786303, "loss": 1.543, "step": 6465 }, { "epoch": 0.64, "learning_rate": 0.0006138898980574923, "loss": 1.5947, "step": 6466 }, { "epoch": 0.64, "learning_rate": 0.0006135953786012391, "loss": 1.4609, "step": 6467 }, { "epoch": 0.64, "learning_rate": 0.0006133008985398974, "loss": 1.416, "step": 6468 }, { "epoch": 0.64, "learning_rate": 0.0006130064579034909, "loss": 1.5439, "step": 6469 }, { "epoch": 0.64, "learning_rate": 0.000612712056722038, "loss": 1.4775, "step": 6470 }, { "epoch": 0.64, "learning_rate": 0.0006124176950255538, "loss": 1.5234, "step": 6471 }, { "epoch": 0.64, "learning_rate": 0.0006121233728440492, "loss": 1.3711, "step": 6472 }, { "epoch": 0.64, "learning_rate": 0.0006118290902075309, "loss": 1.501, "step": 6473 }, { "epoch": 0.64, "learning_rate": 0.0006115348471460021, "loss": 1.4072, "step": 6474 }, { "epoch": 0.64, "learning_rate": 0.0006112406436894612, "loss": 1.4678, "step": 6475 }, { "epoch": 0.64, "learning_rate": 0.0006109464798679032, "loss": 1.5049, "step": 6476 }, { "epoch": 0.64, "learning_rate": 0.0006106523557113187, "loss": 1.6123, "step": 6477 }, { "epoch": 0.64, "learning_rate": 0.0006103582712496947, "loss": 1.4736, "step": 6478 }, { "epoch": 0.64, "learning_rate": 0.0006100642265130136, "loss": 1.5654, "step": 6479 }, { "epoch": 0.64, "learning_rate": 0.0006097702215312542, "loss": 1.5635, "step": 6480 }, { "epoch": 0.64, "learning_rate": 0.0006094762563343906, "loss": 1.5127, "step": 6481 }, { "epoch": 0.64, "learning_rate": 0.0006091823309523937, "loss": 1.4277, "step": 6482 }, { "epoch": 0.64, "learning_rate": 0.0006088884454152298, "loss": 1.4756, "step": 6483 }, { "epoch": 0.64, "learning_rate": 0.0006085945997528613, "loss": 1.5166, "step": 6484 }, { "epoch": 0.64, "learning_rate": 0.0006083007939952463, "loss": 1.4775, "step": 6485 }, { "epoch": 0.64, "learning_rate": 0.0006080070281723388, "loss": 1.6631, "step": 6486 }, { "epoch": 0.64, "learning_rate": 0.00060771330231409, "loss": 1.6631, "step": 6487 }, { "epoch": 0.64, "learning_rate": 0.0006074196164504448, "loss": 1.5195, "step": 6488 }, { "epoch": 0.64, "learning_rate": 0.0006071259706113461, "loss": 1.5332, "step": 6489 }, { "epoch": 0.64, "learning_rate": 0.0006068323648267312, "loss": 1.4268, "step": 6490 }, { "epoch": 0.64, "learning_rate": 0.0006065387991265344, "loss": 1.4443, "step": 6491 }, { "epoch": 0.64, "learning_rate": 0.0006062452735406849, "loss": 1.4902, "step": 6492 }, { "epoch": 0.64, "learning_rate": 0.0006059517880991087, "loss": 1.5537, "step": 6493 }, { "epoch": 0.64, "learning_rate": 0.0006056583428317278, "loss": 1.5361, "step": 6494 }, { "epoch": 0.64, "learning_rate": 0.0006053649377684585, "loss": 1.394, "step": 6495 }, { "epoch": 0.64, "learning_rate": 0.0006050715729392154, "loss": 1.4062, "step": 6496 }, { "epoch": 0.64, "learning_rate": 0.0006047782483739073, "loss": 1.3779, "step": 6497 }, { "epoch": 0.64, "learning_rate": 0.0006044849641024392, "loss": 1.4512, "step": 6498 }, { "epoch": 0.64, "learning_rate": 0.0006041917201547124, "loss": 1.5801, "step": 6499 }, { "epoch": 0.64, "learning_rate": 0.0006038985165606237, "loss": 1.5645, "step": 6500 }, { "epoch": 0.64, "learning_rate": 0.000603605353350066, "loss": 1.4297, "step": 6501 }, { "epoch": 0.64, "learning_rate": 0.0006033122305529279, "loss": 1.5439, "step": 6502 }, { "epoch": 0.64, "learning_rate": 0.0006030191481990943, "loss": 1.4492, "step": 6503 }, { "epoch": 0.64, "learning_rate": 0.0006027261063184453, "loss": 1.4912, "step": 6504 }, { "epoch": 0.64, "learning_rate": 0.0006024331049408575, "loss": 1.4658, "step": 6505 }, { "epoch": 0.64, "learning_rate": 0.0006021401440962031, "loss": 1.5195, "step": 6506 }, { "epoch": 0.64, "learning_rate": 0.0006018472238143502, "loss": 1.4824, "step": 6507 }, { "epoch": 0.64, "learning_rate": 0.0006015543441251625, "loss": 1.5127, "step": 6508 }, { "epoch": 0.64, "learning_rate": 0.0006012615050585002, "loss": 1.5283, "step": 6509 }, { "epoch": 0.64, "learning_rate": 0.0006009687066442186, "loss": 1.4453, "step": 6510 }, { "epoch": 0.64, "learning_rate": 0.0006006759489121696, "loss": 1.417, "step": 6511 }, { "epoch": 0.64, "learning_rate": 0.0006003832318922002, "loss": 1.5889, "step": 6512 }, { "epoch": 0.64, "learning_rate": 0.0006000905556141538, "loss": 1.5146, "step": 6513 }, { "epoch": 0.64, "learning_rate": 0.0005997979201078696, "loss": 1.4355, "step": 6514 }, { "epoch": 0.64, "learning_rate": 0.0005995053254031823, "loss": 1.4258, "step": 6515 }, { "epoch": 0.64, "learning_rate": 0.0005992127715299229, "loss": 1.4551, "step": 6516 }, { "epoch": 0.64, "learning_rate": 0.0005989202585179176, "loss": 1.4775, "step": 6517 }, { "epoch": 0.64, "learning_rate": 0.0005986277863969895, "loss": 1.5068, "step": 6518 }, { "epoch": 0.64, "learning_rate": 0.0005983353551969561, "loss": 1.5596, "step": 6519 }, { "epoch": 0.64, "learning_rate": 0.000598042964947632, "loss": 1.5078, "step": 6520 }, { "epoch": 0.64, "learning_rate": 0.0005977506156788268, "loss": 1.4756, "step": 6521 }, { "epoch": 0.64, "learning_rate": 0.0005974583074203464, "loss": 1.4463, "step": 6522 }, { "epoch": 0.64, "learning_rate": 0.0005971660402019922, "loss": 1.4707, "step": 6523 }, { "epoch": 0.64, "learning_rate": 0.0005968738140535619, "loss": 1.5244, "step": 6524 }, { "epoch": 0.64, "learning_rate": 0.0005965816290048484, "loss": 1.6328, "step": 6525 }, { "epoch": 0.64, "learning_rate": 0.0005962894850856409, "loss": 1.4004, "step": 6526 }, { "epoch": 0.64, "learning_rate": 0.0005959973823257238, "loss": 1.4893, "step": 6527 }, { "epoch": 0.64, "learning_rate": 0.0005957053207548779, "loss": 1.5576, "step": 6528 }, { "epoch": 0.64, "learning_rate": 0.00059541330040288, "loss": 1.5, "step": 6529 }, { "epoch": 0.64, "learning_rate": 0.0005951213212995019, "loss": 1.5322, "step": 6530 }, { "epoch": 0.64, "learning_rate": 0.0005948293834745115, "loss": 1.3931, "step": 6531 }, { "epoch": 0.64, "learning_rate": 0.000594537486957673, "loss": 1.5176, "step": 6532 }, { "epoch": 0.64, "learning_rate": 0.0005942456317787457, "loss": 1.542, "step": 6533 }, { "epoch": 0.64, "learning_rate": 0.0005939538179674848, "loss": 1.5752, "step": 6534 }, { "epoch": 0.64, "learning_rate": 0.0005936620455536418, "loss": 1.4307, "step": 6535 }, { "epoch": 0.64, "learning_rate": 0.0005933703145669635, "loss": 1.5898, "step": 6536 }, { "epoch": 0.64, "learning_rate": 0.0005930786250371923, "loss": 1.4775, "step": 6537 }, { "epoch": 0.64, "learning_rate": 0.000592786976994067, "loss": 1.4404, "step": 6538 }, { "epoch": 0.64, "learning_rate": 0.0005924953704673219, "loss": 1.4854, "step": 6539 }, { "epoch": 0.64, "learning_rate": 0.0005922038054866869, "loss": 1.5156, "step": 6540 }, { "epoch": 0.64, "learning_rate": 0.0005919122820818878, "loss": 1.5352, "step": 6541 }, { "epoch": 0.64, "learning_rate": 0.000591620800282646, "loss": 1.4209, "step": 6542 }, { "epoch": 0.65, "learning_rate": 0.0005913293601186788, "loss": 1.5215, "step": 6543 }, { "epoch": 0.65, "learning_rate": 0.0005910379616196995, "loss": 1.416, "step": 6544 }, { "epoch": 0.65, "learning_rate": 0.0005907466048154169, "loss": 1.4453, "step": 6545 }, { "epoch": 0.65, "learning_rate": 0.0005904552897355349, "loss": 1.3604, "step": 6546 }, { "epoch": 0.65, "learning_rate": 0.0005901640164097546, "loss": 1.6602, "step": 6547 }, { "epoch": 0.65, "learning_rate": 0.0005898727848677718, "loss": 1.5137, "step": 6548 }, { "epoch": 0.65, "learning_rate": 0.000589581595139278, "loss": 1.543, "step": 6549 }, { "epoch": 0.65, "learning_rate": 0.0005892904472539611, "loss": 1.3438, "step": 6550 }, { "epoch": 0.65, "learning_rate": 0.0005889993412415042, "loss": 1.5518, "step": 6551 }, { "epoch": 0.65, "learning_rate": 0.0005887082771315862, "loss": 1.3257, "step": 6552 }, { "epoch": 0.65, "learning_rate": 0.0005884172549538817, "loss": 1.46, "step": 6553 }, { "epoch": 0.65, "learning_rate": 0.0005881262747380615, "loss": 1.4453, "step": 6554 }, { "epoch": 0.65, "learning_rate": 0.0005878353365137914, "loss": 1.5107, "step": 6555 }, { "epoch": 0.65, "learning_rate": 0.0005875444403107335, "loss": 1.2539, "step": 6556 }, { "epoch": 0.65, "learning_rate": 0.0005872535861585456, "loss": 1.4424, "step": 6557 }, { "epoch": 0.65, "learning_rate": 0.0005869627740868803, "loss": 1.4189, "step": 6558 }, { "epoch": 0.65, "learning_rate": 0.0005866720041253873, "loss": 1.4043, "step": 6559 }, { "epoch": 0.65, "learning_rate": 0.0005863812763037111, "loss": 1.3418, "step": 6560 }, { "epoch": 0.65, "learning_rate": 0.000586090590651492, "loss": 1.4688, "step": 6561 }, { "epoch": 0.65, "learning_rate": 0.0005857999471983662, "loss": 1.5488, "step": 6562 }, { "epoch": 0.65, "learning_rate": 0.0005855093459739655, "loss": 1.4346, "step": 6563 }, { "epoch": 0.65, "learning_rate": 0.0005852187870079173, "loss": 1.4902, "step": 6564 }, { "epoch": 0.65, "learning_rate": 0.0005849282703298452, "loss": 1.4062, "step": 6565 }, { "epoch": 0.65, "learning_rate": 0.0005846377959693678, "loss": 1.4175, "step": 6566 }, { "epoch": 0.65, "learning_rate": 0.0005843473639560998, "loss": 1.541, "step": 6567 }, { "epoch": 0.65, "learning_rate": 0.0005840569743196513, "loss": 1.457, "step": 6568 }, { "epoch": 0.65, "learning_rate": 0.0005837666270896285, "loss": 1.4717, "step": 6569 }, { "epoch": 0.65, "learning_rate": 0.0005834763222956328, "loss": 1.4307, "step": 6570 }, { "epoch": 0.65, "learning_rate": 0.0005831860599672616, "loss": 1.5615, "step": 6571 }, { "epoch": 0.65, "learning_rate": 0.0005828958401341078, "loss": 1.5439, "step": 6572 }, { "epoch": 0.65, "learning_rate": 0.0005826056628257601, "loss": 1.373, "step": 6573 }, { "epoch": 0.65, "learning_rate": 0.0005823155280718027, "loss": 1.4951, "step": 6574 }, { "epoch": 0.65, "learning_rate": 0.0005820254359018156, "loss": 1.4297, "step": 6575 }, { "epoch": 0.65, "learning_rate": 0.0005817353863453745, "loss": 1.3633, "step": 6576 }, { "epoch": 0.65, "learning_rate": 0.0005814453794320507, "loss": 1.4258, "step": 6577 }, { "epoch": 0.65, "learning_rate": 0.000581155415191411, "loss": 1.3877, "step": 6578 }, { "epoch": 0.65, "learning_rate": 0.000580865493653018, "loss": 1.5176, "step": 6579 }, { "epoch": 0.65, "learning_rate": 0.0005805756148464298, "loss": 1.4463, "step": 6580 }, { "epoch": 0.65, "learning_rate": 0.0005802857788012007, "loss": 1.459, "step": 6581 }, { "epoch": 0.65, "learning_rate": 0.0005799959855468797, "loss": 1.417, "step": 6582 }, { "epoch": 0.65, "learning_rate": 0.0005797062351130122, "loss": 1.3975, "step": 6583 }, { "epoch": 0.65, "learning_rate": 0.000579416527529139, "loss": 1.5264, "step": 6584 }, { "epoch": 0.65, "learning_rate": 0.0005791268628247963, "loss": 1.4111, "step": 6585 }, { "epoch": 0.65, "learning_rate": 0.0005788372410295165, "loss": 1.4258, "step": 6586 }, { "epoch": 0.65, "learning_rate": 0.0005785476621728269, "loss": 1.418, "step": 6587 }, { "epoch": 0.65, "learning_rate": 0.0005782581262842509, "loss": 1.5576, "step": 6588 }, { "epoch": 0.65, "learning_rate": 0.0005779686333933074, "loss": 1.4404, "step": 6589 }, { "epoch": 0.65, "learning_rate": 0.000577679183529511, "loss": 1.4902, "step": 6590 }, { "epoch": 0.65, "learning_rate": 0.0005773897767223715, "loss": 1.4551, "step": 6591 }, { "epoch": 0.65, "learning_rate": 0.000577100413001395, "loss": 1.4834, "step": 6592 }, { "epoch": 0.65, "learning_rate": 0.0005768110923960831, "loss": 1.5146, "step": 6593 }, { "epoch": 0.65, "learning_rate": 0.0005765218149359324, "loss": 1.3672, "step": 6594 }, { "epoch": 0.65, "learning_rate": 0.0005762325806504353, "loss": 1.5605, "step": 6595 }, { "epoch": 0.65, "learning_rate": 0.0005759433895690802, "loss": 1.6201, "step": 6596 }, { "epoch": 0.65, "learning_rate": 0.0005756542417213508, "loss": 1.4863, "step": 6597 }, { "epoch": 0.65, "learning_rate": 0.0005753651371367266, "loss": 1.4023, "step": 6598 }, { "epoch": 0.65, "learning_rate": 0.0005750760758446821, "loss": 1.5684, "step": 6599 }, { "epoch": 0.65, "learning_rate": 0.0005747870578746882, "loss": 1.3887, "step": 6600 }, { "epoch": 0.65, "learning_rate": 0.0005744980832562108, "loss": 1.377, "step": 6601 }, { "epoch": 0.65, "learning_rate": 0.0005742091520187119, "loss": 1.5352, "step": 6602 }, { "epoch": 0.65, "learning_rate": 0.0005739202641916484, "loss": 1.3682, "step": 6603 }, { "epoch": 0.65, "learning_rate": 0.0005736314198044731, "loss": 1.5967, "step": 6604 }, { "epoch": 0.65, "learning_rate": 0.0005733426188866347, "loss": 1.5322, "step": 6605 }, { "epoch": 0.65, "learning_rate": 0.0005730538614675769, "loss": 1.4434, "step": 6606 }, { "epoch": 0.65, "learning_rate": 0.0005727651475767395, "loss": 1.5488, "step": 6607 }, { "epoch": 0.65, "learning_rate": 0.0005724764772435574, "loss": 1.4443, "step": 6608 }, { "epoch": 0.65, "learning_rate": 0.0005721878504974609, "loss": 1.4248, "step": 6609 }, { "epoch": 0.65, "learning_rate": 0.0005718992673678771, "loss": 1.3262, "step": 6610 }, { "epoch": 0.65, "learning_rate": 0.0005716107278842273, "loss": 1.3887, "step": 6611 }, { "epoch": 0.65, "learning_rate": 0.0005713222320759288, "loss": 1.4766, "step": 6612 }, { "epoch": 0.65, "learning_rate": 0.0005710337799723945, "loss": 1.4395, "step": 6613 }, { "epoch": 0.65, "learning_rate": 0.0005707453716030327, "loss": 1.3369, "step": 6614 }, { "epoch": 0.65, "learning_rate": 0.0005704570069972475, "loss": 1.3252, "step": 6615 }, { "epoch": 0.65, "learning_rate": 0.0005701686861844384, "loss": 1.3662, "step": 6616 }, { "epoch": 0.65, "learning_rate": 0.0005698804091940004, "loss": 1.4238, "step": 6617 }, { "epoch": 0.65, "learning_rate": 0.0005695921760553239, "loss": 1.5254, "step": 6618 }, { "epoch": 0.65, "learning_rate": 0.000569303986797795, "loss": 1.418, "step": 6619 }, { "epoch": 0.65, "learning_rate": 0.0005690158414507955, "loss": 1.4258, "step": 6620 }, { "epoch": 0.65, "learning_rate": 0.0005687277400437025, "loss": 1.4121, "step": 6621 }, { "epoch": 0.65, "learning_rate": 0.0005684396826058886, "loss": 1.4321, "step": 6622 }, { "epoch": 0.65, "learning_rate": 0.0005681516691667218, "loss": 1.2959, "step": 6623 }, { "epoch": 0.65, "learning_rate": 0.0005678636997555663, "loss": 1.5381, "step": 6624 }, { "epoch": 0.65, "learning_rate": 0.0005675757744017806, "loss": 1.3062, "step": 6625 }, { "epoch": 0.65, "learning_rate": 0.00056728789313472, "loss": 1.3516, "step": 6626 }, { "epoch": 0.65, "learning_rate": 0.000567000055983734, "loss": 1.4219, "step": 6627 }, { "epoch": 0.65, "learning_rate": 0.0005667122629781691, "loss": 1.4385, "step": 6628 }, { "epoch": 0.65, "learning_rate": 0.0005664245141473663, "loss": 1.5186, "step": 6629 }, { "epoch": 0.65, "learning_rate": 0.0005661368095206622, "loss": 1.4092, "step": 6630 }, { "epoch": 0.65, "learning_rate": 0.0005658491491273894, "loss": 1.4395, "step": 6631 }, { "epoch": 0.65, "learning_rate": 0.0005655615329968746, "loss": 1.4238, "step": 6632 }, { "epoch": 0.65, "learning_rate": 0.0005652739611584418, "loss": 1.3887, "step": 6633 }, { "epoch": 0.65, "learning_rate": 0.0005649864336414094, "loss": 1.2959, "step": 6634 }, { "epoch": 0.65, "learning_rate": 0.0005646989504750915, "loss": 1.5166, "step": 6635 }, { "epoch": 0.65, "learning_rate": 0.000564411511688798, "loss": 1.5742, "step": 6636 }, { "epoch": 0.65, "learning_rate": 0.0005641241173118337, "loss": 1.4248, "step": 6637 }, { "epoch": 0.65, "learning_rate": 0.0005638367673734992, "loss": 1.4619, "step": 6638 }, { "epoch": 0.65, "learning_rate": 0.0005635494619030907, "loss": 1.6367, "step": 6639 }, { "epoch": 0.65, "learning_rate": 0.0005632622009298996, "loss": 1.3369, "step": 6640 }, { "epoch": 0.65, "learning_rate": 0.000562974984483213, "loss": 1.3428, "step": 6641 }, { "epoch": 0.65, "learning_rate": 0.0005626878125923127, "loss": 1.6064, "step": 6642 }, { "epoch": 0.65, "learning_rate": 0.0005624006852864776, "loss": 1.4209, "step": 6643 }, { "epoch": 0.65, "learning_rate": 0.0005621136025949805, "loss": 1.4922, "step": 6644 }, { "epoch": 0.66, "learning_rate": 0.0005618265645470901, "loss": 1.5674, "step": 6645 }, { "epoch": 0.66, "learning_rate": 0.0005615395711720711, "loss": 1.4492, "step": 6646 }, { "epoch": 0.66, "learning_rate": 0.0005612526224991828, "loss": 1.4375, "step": 6647 }, { "epoch": 0.66, "learning_rate": 0.0005609657185576803, "loss": 1.4766, "step": 6648 }, { "epoch": 0.66, "learning_rate": 0.0005606788593768144, "loss": 1.4932, "step": 6649 }, { "epoch": 0.66, "learning_rate": 0.0005603920449858309, "loss": 1.3555, "step": 6650 }, { "epoch": 0.66, "learning_rate": 0.0005601052754139714, "loss": 1.3281, "step": 6651 }, { "epoch": 0.66, "learning_rate": 0.0005598185506904727, "loss": 1.4482, "step": 6652 }, { "epoch": 0.66, "learning_rate": 0.0005595318708445672, "loss": 1.4619, "step": 6653 }, { "epoch": 0.66, "learning_rate": 0.0005592452359054823, "loss": 1.4473, "step": 6654 }, { "epoch": 0.66, "learning_rate": 0.0005589586459024415, "loss": 1.458, "step": 6655 }, { "epoch": 0.66, "learning_rate": 0.0005586721008646632, "loss": 1.5029, "step": 6656 }, { "epoch": 0.66, "learning_rate": 0.0005583856008213615, "loss": 1.4297, "step": 6657 }, { "epoch": 0.66, "learning_rate": 0.0005580991458017455, "loss": 1.4131, "step": 6658 }, { "epoch": 0.66, "learning_rate": 0.0005578127358350205, "loss": 1.5908, "step": 6659 }, { "epoch": 0.66, "learning_rate": 0.0005575263709503857, "loss": 1.4238, "step": 6660 }, { "epoch": 0.66, "learning_rate": 0.0005572400511770381, "loss": 1.5684, "step": 6661 }, { "epoch": 0.66, "learning_rate": 0.000556953776544168, "loss": 1.4844, "step": 6662 }, { "epoch": 0.66, "learning_rate": 0.0005566675470809615, "loss": 1.4062, "step": 6663 }, { "epoch": 0.66, "learning_rate": 0.000556381362816601, "loss": 1.5107, "step": 6664 }, { "epoch": 0.66, "learning_rate": 0.0005560952237802636, "loss": 1.4639, "step": 6665 }, { "epoch": 0.66, "learning_rate": 0.0005558091300011215, "loss": 1.502, "step": 6666 }, { "epoch": 0.66, "learning_rate": 0.0005555230815083431, "loss": 1.4434, "step": 6667 }, { "epoch": 0.66, "learning_rate": 0.0005552370783310914, "loss": 1.4976, "step": 6668 }, { "epoch": 0.66, "learning_rate": 0.0005549511204985254, "loss": 1.4507, "step": 6669 }, { "epoch": 0.66, "learning_rate": 0.0005546652080397991, "loss": 1.3809, "step": 6670 }, { "epoch": 0.66, "learning_rate": 0.000554379340984062, "loss": 1.5674, "step": 6671 }, { "epoch": 0.66, "learning_rate": 0.0005540935193604588, "loss": 1.5039, "step": 6672 }, { "epoch": 0.66, "learning_rate": 0.0005538077431981301, "loss": 1.4854, "step": 6673 }, { "epoch": 0.66, "learning_rate": 0.0005535220125262113, "loss": 1.3965, "step": 6674 }, { "epoch": 0.66, "learning_rate": 0.0005532363273738331, "loss": 1.4404, "step": 6675 }, { "epoch": 0.66, "learning_rate": 0.0005529506877701222, "loss": 1.416, "step": 6676 }, { "epoch": 0.66, "learning_rate": 0.0005526650937441999, "loss": 1.3877, "step": 6677 }, { "epoch": 0.66, "learning_rate": 0.0005523795453251832, "loss": 1.5645, "step": 6678 }, { "epoch": 0.66, "learning_rate": 0.0005520940425421851, "loss": 1.4707, "step": 6679 }, { "epoch": 0.66, "learning_rate": 0.0005518085854243129, "loss": 1.458, "step": 6680 }, { "epoch": 0.66, "learning_rate": 0.0005515231740006696, "loss": 1.4053, "step": 6681 }, { "epoch": 0.66, "learning_rate": 0.0005512378083003537, "loss": 1.4316, "step": 6682 }, { "epoch": 0.66, "learning_rate": 0.0005509524883524589, "loss": 1.5889, "step": 6683 }, { "epoch": 0.66, "learning_rate": 0.0005506672141860744, "loss": 1.4424, "step": 6684 }, { "epoch": 0.66, "learning_rate": 0.0005503819858302844, "loss": 1.5957, "step": 6685 }, { "epoch": 0.66, "learning_rate": 0.0005500968033141688, "loss": 1.4814, "step": 6686 }, { "epoch": 0.66, "learning_rate": 0.0005498116666668029, "loss": 1.5381, "step": 6687 }, { "epoch": 0.66, "learning_rate": 0.0005495265759172566, "loss": 1.5557, "step": 6688 }, { "epoch": 0.66, "learning_rate": 0.0005492415310945957, "loss": 1.4639, "step": 6689 }, { "epoch": 0.66, "learning_rate": 0.0005489565322278814, "loss": 1.4688, "step": 6690 }, { "epoch": 0.66, "learning_rate": 0.00054867157934617, "loss": 1.3633, "step": 6691 }, { "epoch": 0.66, "learning_rate": 0.0005483866724785128, "loss": 1.5273, "step": 6692 }, { "epoch": 0.66, "learning_rate": 0.0005481018116539573, "loss": 1.46, "step": 6693 }, { "epoch": 0.66, "learning_rate": 0.0005478169969015457, "loss": 1.311, "step": 6694 }, { "epoch": 0.66, "learning_rate": 0.0005475322282503154, "loss": 1.5859, "step": 6695 }, { "epoch": 0.66, "learning_rate": 0.0005472475057292994, "loss": 1.5244, "step": 6696 }, { "epoch": 0.66, "learning_rate": 0.0005469628293675257, "loss": 1.4229, "step": 6697 }, { "epoch": 0.66, "learning_rate": 0.0005466781991940179, "loss": 1.5459, "step": 6698 }, { "epoch": 0.66, "learning_rate": 0.0005463936152377949, "loss": 1.3506, "step": 6699 }, { "epoch": 0.66, "learning_rate": 0.0005461090775278702, "loss": 1.3887, "step": 6700 }, { "epoch": 0.66, "learning_rate": 0.0005458245860932536, "loss": 1.5176, "step": 6701 }, { "epoch": 0.66, "learning_rate": 0.0005455401409629497, "loss": 1.333, "step": 6702 }, { "epoch": 0.66, "learning_rate": 0.000545255742165958, "loss": 1.3955, "step": 6703 }, { "epoch": 0.66, "learning_rate": 0.0005449713897312743, "loss": 1.4434, "step": 6704 }, { "epoch": 0.66, "learning_rate": 0.0005446870836878884, "loss": 1.4404, "step": 6705 }, { "epoch": 0.66, "learning_rate": 0.0005444028240647865, "loss": 1.5488, "step": 6706 }, { "epoch": 0.66, "learning_rate": 0.0005441186108909492, "loss": 1.4893, "step": 6707 }, { "epoch": 0.66, "learning_rate": 0.0005438344441953529, "loss": 1.4775, "step": 6708 }, { "epoch": 0.66, "learning_rate": 0.000543550324006969, "loss": 1.4863, "step": 6709 }, { "epoch": 0.66, "learning_rate": 0.0005432662503547642, "loss": 1.4316, "step": 6710 }, { "epoch": 0.66, "learning_rate": 0.0005429822232677009, "loss": 1.4551, "step": 6711 }, { "epoch": 0.66, "learning_rate": 0.0005426982427747363, "loss": 1.5439, "step": 6712 }, { "epoch": 0.66, "learning_rate": 0.0005424143089048226, "loss": 1.4014, "step": 6713 }, { "epoch": 0.66, "learning_rate": 0.0005421304216869078, "loss": 1.46, "step": 6714 }, { "epoch": 0.66, "learning_rate": 0.0005418465811499347, "loss": 1.4561, "step": 6715 }, { "epoch": 0.66, "learning_rate": 0.0005415627873228418, "loss": 1.418, "step": 6716 }, { "epoch": 0.66, "learning_rate": 0.0005412790402345624, "loss": 1.3789, "step": 6717 }, { "epoch": 0.66, "learning_rate": 0.0005409953399140253, "loss": 1.3682, "step": 6718 }, { "epoch": 0.66, "learning_rate": 0.0005407116863901545, "loss": 1.4795, "step": 6719 }, { "epoch": 0.66, "learning_rate": 0.000540428079691869, "loss": 1.3896, "step": 6720 }, { "epoch": 0.66, "learning_rate": 0.0005401445198480834, "loss": 1.5547, "step": 6721 }, { "epoch": 0.66, "learning_rate": 0.0005398610068877073, "loss": 1.3623, "step": 6722 }, { "epoch": 0.66, "learning_rate": 0.0005395775408396455, "loss": 1.4058, "step": 6723 }, { "epoch": 0.66, "learning_rate": 0.000539294121732798, "loss": 1.5312, "step": 6724 }, { "epoch": 0.66, "learning_rate": 0.0005390107495960604, "loss": 1.4062, "step": 6725 }, { "epoch": 0.66, "learning_rate": 0.0005387274244583227, "loss": 1.4722, "step": 6726 }, { "epoch": 0.66, "learning_rate": 0.0005384441463484708, "loss": 1.4834, "step": 6727 }, { "epoch": 0.66, "learning_rate": 0.0005381609152953855, "loss": 1.4707, "step": 6728 }, { "epoch": 0.66, "learning_rate": 0.0005378777313279433, "loss": 1.4072, "step": 6729 }, { "epoch": 0.66, "learning_rate": 0.0005375945944750154, "loss": 1.4062, "step": 6730 }, { "epoch": 0.66, "learning_rate": 0.0005373115047654681, "loss": 1.4023, "step": 6731 }, { "epoch": 0.66, "learning_rate": 0.0005370284622281631, "loss": 1.5049, "step": 6732 }, { "epoch": 0.66, "learning_rate": 0.0005367454668919575, "loss": 1.3682, "step": 6733 }, { "epoch": 0.66, "learning_rate": 0.0005364625187857033, "loss": 1.5391, "step": 6734 }, { "epoch": 0.66, "learning_rate": 0.0005361796179382477, "loss": 1.3535, "step": 6735 }, { "epoch": 0.66, "learning_rate": 0.0005358967643784332, "loss": 1.4272, "step": 6736 }, { "epoch": 0.66, "learning_rate": 0.0005356139581350973, "loss": 1.6689, "step": 6737 }, { "epoch": 0.66, "learning_rate": 0.0005353311992370728, "loss": 1.4639, "step": 6738 }, { "epoch": 0.66, "learning_rate": 0.0005350484877131881, "loss": 1.333, "step": 6739 }, { "epoch": 0.66, "learning_rate": 0.0005347658235922657, "loss": 1.5254, "step": 6740 }, { "epoch": 0.66, "learning_rate": 0.0005344832069031243, "loss": 1.5146, "step": 6741 }, { "epoch": 0.66, "learning_rate": 0.0005342006376745772, "loss": 1.5107, "step": 6742 }, { "epoch": 0.66, "learning_rate": 0.0005339181159354331, "loss": 1.4277, "step": 6743 }, { "epoch": 0.66, "learning_rate": 0.0005336356417144961, "loss": 1.3516, "step": 6744 }, { "epoch": 0.66, "learning_rate": 0.0005333532150405646, "loss": 1.54, "step": 6745 }, { "epoch": 0.67, "learning_rate": 0.0005330708359424331, "loss": 1.4473, "step": 6746 }, { "epoch": 0.67, "learning_rate": 0.0005327885044488907, "loss": 1.3779, "step": 6747 }, { "epoch": 0.67, "learning_rate": 0.0005325062205887216, "loss": 1.5645, "step": 6748 }, { "epoch": 0.67, "learning_rate": 0.000532223984390706, "loss": 1.3916, "step": 6749 }, { "epoch": 0.67, "learning_rate": 0.0005319417958836178, "loss": 1.3735, "step": 6750 }, { "epoch": 0.67, "learning_rate": 0.0005316596550962271, "loss": 1.5166, "step": 6751 }, { "epoch": 0.67, "learning_rate": 0.0005313775620572991, "loss": 1.5469, "step": 6752 }, { "epoch": 0.67, "learning_rate": 0.0005310955167955937, "loss": 1.5049, "step": 6753 }, { "epoch": 0.67, "learning_rate": 0.0005308135193398661, "loss": 1.4849, "step": 6754 }, { "epoch": 0.67, "learning_rate": 0.0005305315697188666, "loss": 1.5674, "step": 6755 }, { "epoch": 0.67, "learning_rate": 0.0005302496679613409, "loss": 1.583, "step": 6756 }, { "epoch": 0.67, "learning_rate": 0.0005299678140960291, "loss": 1.4395, "step": 6757 }, { "epoch": 0.67, "learning_rate": 0.0005296860081516674, "loss": 1.4561, "step": 6758 }, { "epoch": 0.67, "learning_rate": 0.0005294042501569864, "loss": 1.5068, "step": 6759 }, { "epoch": 0.67, "learning_rate": 0.0005291225401407117, "loss": 1.4873, "step": 6760 }, { "epoch": 0.67, "learning_rate": 0.000528840878131565, "loss": 1.4863, "step": 6761 }, { "epoch": 0.67, "learning_rate": 0.0005285592641582622, "loss": 1.3301, "step": 6762 }, { "epoch": 0.67, "learning_rate": 0.0005282776982495146, "loss": 1.4678, "step": 6763 }, { "epoch": 0.67, "learning_rate": 0.0005279961804340282, "loss": 1.4336, "step": 6764 }, { "epoch": 0.67, "learning_rate": 0.0005277147107405047, "loss": 1.4209, "step": 6765 }, { "epoch": 0.67, "learning_rate": 0.0005274332891976407, "loss": 1.4941, "step": 6766 }, { "epoch": 0.67, "learning_rate": 0.0005271519158341276, "loss": 1.542, "step": 6767 }, { "epoch": 0.67, "learning_rate": 0.0005268705906786524, "loss": 1.3496, "step": 6768 }, { "epoch": 0.67, "learning_rate": 0.0005265893137598965, "loss": 1.4531, "step": 6769 }, { "epoch": 0.67, "learning_rate": 0.0005263080851065371, "loss": 1.3896, "step": 6770 }, { "epoch": 0.67, "learning_rate": 0.0005260269047472464, "loss": 1.375, "step": 6771 }, { "epoch": 0.67, "learning_rate": 0.0005257457727106907, "loss": 1.4795, "step": 6772 }, { "epoch": 0.67, "learning_rate": 0.0005254646890255328, "loss": 1.3545, "step": 6773 }, { "epoch": 0.67, "learning_rate": 0.0005251836537204294, "loss": 1.3618, "step": 6774 }, { "epoch": 0.67, "learning_rate": 0.0005249026668240331, "loss": 1.5957, "step": 6775 }, { "epoch": 0.67, "learning_rate": 0.000524621728364991, "loss": 1.4443, "step": 6776 }, { "epoch": 0.67, "learning_rate": 0.0005243408383719454, "loss": 1.5254, "step": 6777 }, { "epoch": 0.67, "learning_rate": 0.0005240599968735338, "loss": 1.5537, "step": 6778 }, { "epoch": 0.67, "learning_rate": 0.000523779203898389, "loss": 1.3887, "step": 6779 }, { "epoch": 0.67, "learning_rate": 0.0005234984594751382, "loss": 1.3965, "step": 6780 }, { "epoch": 0.67, "learning_rate": 0.0005232177636324042, "loss": 1.5264, "step": 6781 }, { "epoch": 0.67, "learning_rate": 0.0005229371163988046, "loss": 1.4229, "step": 6782 }, { "epoch": 0.67, "learning_rate": 0.0005226565178029519, "loss": 1.376, "step": 6783 }, { "epoch": 0.67, "learning_rate": 0.000522375967873454, "loss": 1.4375, "step": 6784 }, { "epoch": 0.67, "learning_rate": 0.0005220954666389136, "loss": 1.5225, "step": 6785 }, { "epoch": 0.67, "learning_rate": 0.0005218150141279286, "loss": 1.4199, "step": 6786 }, { "epoch": 0.67, "learning_rate": 0.0005215346103690915, "loss": 1.4321, "step": 6787 }, { "epoch": 0.67, "learning_rate": 0.0005212542553909905, "loss": 1.4219, "step": 6788 }, { "epoch": 0.67, "learning_rate": 0.0005209739492222085, "loss": 1.3721, "step": 6789 }, { "epoch": 0.67, "learning_rate": 0.0005206936918913232, "loss": 1.3389, "step": 6790 }, { "epoch": 0.67, "learning_rate": 0.0005204134834269076, "loss": 1.394, "step": 6791 }, { "epoch": 0.67, "learning_rate": 0.0005201333238575298, "loss": 1.4834, "step": 6792 }, { "epoch": 0.67, "learning_rate": 0.0005198532132117525, "loss": 1.2861, "step": 6793 }, { "epoch": 0.67, "learning_rate": 0.0005195731515181339, "loss": 1.4326, "step": 6794 }, { "epoch": 0.67, "learning_rate": 0.0005192931388052269, "loss": 1.4844, "step": 6795 }, { "epoch": 0.67, "learning_rate": 0.000519013175101579, "loss": 1.4414, "step": 6796 }, { "epoch": 0.67, "learning_rate": 0.0005187332604357342, "loss": 1.4062, "step": 6797 }, { "epoch": 0.67, "learning_rate": 0.0005184533948362299, "loss": 1.4443, "step": 6798 }, { "epoch": 0.67, "learning_rate": 0.0005181735783315997, "loss": 1.5967, "step": 6799 }, { "epoch": 0.67, "learning_rate": 0.0005178938109503706, "loss": 1.5059, "step": 6800 }, { "epoch": 0.67, "learning_rate": 0.000517614092721066, "loss": 1.5078, "step": 6801 }, { "epoch": 0.67, "learning_rate": 0.0005173344236722039, "loss": 1.4053, "step": 6802 }, { "epoch": 0.67, "learning_rate": 0.0005170548038322974, "loss": 1.4424, "step": 6803 }, { "epoch": 0.67, "learning_rate": 0.0005167752332298541, "loss": 1.415, "step": 6804 }, { "epoch": 0.67, "learning_rate": 0.0005164957118933773, "loss": 1.3809, "step": 6805 }, { "epoch": 0.67, "learning_rate": 0.0005162162398513648, "loss": 1.3975, "step": 6806 }, { "epoch": 0.67, "learning_rate": 0.0005159368171323091, "loss": 1.4932, "step": 6807 }, { "epoch": 0.67, "learning_rate": 0.0005156574437646986, "loss": 1.3975, "step": 6808 }, { "epoch": 0.67, "learning_rate": 0.0005153781197770157, "loss": 1.5303, "step": 6809 }, { "epoch": 0.67, "learning_rate": 0.000515098845197738, "loss": 1.3086, "step": 6810 }, { "epoch": 0.67, "learning_rate": 0.000514819620055339, "loss": 1.3711, "step": 6811 }, { "epoch": 0.67, "learning_rate": 0.0005145404443782859, "loss": 1.542, "step": 6812 }, { "epoch": 0.67, "learning_rate": 0.0005142613181950413, "loss": 1.373, "step": 6813 }, { "epoch": 0.67, "learning_rate": 0.000513982241534063, "loss": 1.499, "step": 6814 }, { "epoch": 0.67, "learning_rate": 0.0005137032144238035, "loss": 1.4385, "step": 6815 }, { "epoch": 0.67, "learning_rate": 0.0005134242368927103, "loss": 1.4629, "step": 6816 }, { "epoch": 0.67, "learning_rate": 0.0005131453089692258, "loss": 1.5029, "step": 6817 }, { "epoch": 0.67, "learning_rate": 0.0005128664306817874, "loss": 1.4951, "step": 6818 }, { "epoch": 0.67, "learning_rate": 0.0005125876020588275, "loss": 1.417, "step": 6819 }, { "epoch": 0.67, "learning_rate": 0.0005123088231287733, "loss": 1.5654, "step": 6820 }, { "epoch": 0.67, "learning_rate": 0.0005120300939200471, "loss": 1.4375, "step": 6821 }, { "epoch": 0.67, "learning_rate": 0.000511751414461066, "loss": 1.4805, "step": 6822 }, { "epoch": 0.67, "learning_rate": 0.0005114727847802419, "loss": 1.6006, "step": 6823 }, { "epoch": 0.67, "learning_rate": 0.0005111942049059821, "loss": 1.4424, "step": 6824 }, { "epoch": 0.67, "learning_rate": 0.0005109156748666882, "loss": 1.5127, "step": 6825 }, { "epoch": 0.67, "learning_rate": 0.0005106371946907572, "loss": 1.5176, "step": 6826 }, { "epoch": 0.67, "learning_rate": 0.0005103587644065808, "loss": 1.498, "step": 6827 }, { "epoch": 0.67, "learning_rate": 0.0005100803840425454, "loss": 1.4473, "step": 6828 }, { "epoch": 0.67, "learning_rate": 0.0005098020536270331, "loss": 1.4297, "step": 6829 }, { "epoch": 0.67, "learning_rate": 0.0005095237731884202, "loss": 1.6045, "step": 6830 }, { "epoch": 0.67, "learning_rate": 0.000509245542755078, "loss": 1.4434, "step": 6831 }, { "epoch": 0.67, "learning_rate": 0.0005089673623553727, "loss": 1.3535, "step": 6832 }, { "epoch": 0.67, "learning_rate": 0.0005086892320176658, "loss": 1.5518, "step": 6833 }, { "epoch": 0.67, "learning_rate": 0.0005084111517703131, "loss": 1.5117, "step": 6834 }, { "epoch": 0.67, "learning_rate": 0.0005081331216416655, "loss": 1.5684, "step": 6835 }, { "epoch": 0.67, "learning_rate": 0.000507855141660069, "loss": 1.4111, "step": 6836 }, { "epoch": 0.67, "learning_rate": 0.0005075772118538644, "loss": 1.5068, "step": 6837 }, { "epoch": 0.67, "learning_rate": 0.0005072993322513872, "loss": 1.4072, "step": 6838 }, { "epoch": 0.67, "learning_rate": 0.0005070215028809682, "loss": 1.4746, "step": 6839 }, { "epoch": 0.67, "learning_rate": 0.0005067437237709322, "loss": 1.4082, "step": 6840 }, { "epoch": 0.67, "learning_rate": 0.0005064659949496002, "loss": 1.3364, "step": 6841 }, { "epoch": 0.67, "learning_rate": 0.0005061883164452869, "loss": 1.4805, "step": 6842 }, { "epoch": 0.67, "learning_rate": 0.0005059106882863023, "loss": 1.3115, "step": 6843 }, { "epoch": 0.67, "learning_rate": 0.0005056331105009515, "loss": 1.4443, "step": 6844 }, { "epoch": 0.67, "learning_rate": 0.0005053555831175343, "loss": 1.4248, "step": 6845 }, { "epoch": 0.67, "learning_rate": 0.0005050781061643446, "loss": 1.4863, "step": 6846 }, { "epoch": 0.67, "learning_rate": 0.0005048006796696727, "loss": 1.4531, "step": 6847 }, { "epoch": 0.68, "learning_rate": 0.0005045233036618026, "loss": 1.5225, "step": 6848 }, { "epoch": 0.68, "learning_rate": 0.0005042459781690137, "loss": 1.4395, "step": 6849 }, { "epoch": 0.68, "learning_rate": 0.0005039687032195799, "loss": 1.5059, "step": 6850 }, { "epoch": 0.68, "learning_rate": 0.0005036914788417699, "loss": 1.6514, "step": 6851 }, { "epoch": 0.68, "learning_rate": 0.0005034143050638473, "loss": 1.3486, "step": 6852 }, { "epoch": 0.68, "learning_rate": 0.0005031371819140711, "loss": 1.4531, "step": 6853 }, { "epoch": 0.68, "learning_rate": 0.0005028601094206943, "loss": 1.3242, "step": 6854 }, { "epoch": 0.68, "learning_rate": 0.0005025830876119659, "loss": 1.3926, "step": 6855 }, { "epoch": 0.68, "learning_rate": 0.0005023061165161279, "loss": 1.542, "step": 6856 }, { "epoch": 0.68, "learning_rate": 0.0005020291961614186, "loss": 1.4048, "step": 6857 }, { "epoch": 0.68, "learning_rate": 0.0005017523265760708, "loss": 1.4043, "step": 6858 }, { "epoch": 0.68, "learning_rate": 0.000501475507788312, "loss": 1.4854, "step": 6859 }, { "epoch": 0.68, "learning_rate": 0.0005011987398263646, "loss": 1.5244, "step": 6860 }, { "epoch": 0.68, "learning_rate": 0.0005009220227184452, "loss": 1.5117, "step": 6861 }, { "epoch": 0.68, "learning_rate": 0.000500645356492767, "loss": 1.3115, "step": 6862 }, { "epoch": 0.68, "learning_rate": 0.000500368741177536, "loss": 1.4297, "step": 6863 }, { "epoch": 0.68, "learning_rate": 0.0005000921768009541, "loss": 1.542, "step": 6864 }, { "epoch": 0.68, "learning_rate": 0.0004998156633912174, "loss": 1.541, "step": 6865 }, { "epoch": 0.68, "learning_rate": 0.0004995392009765175, "loss": 1.4346, "step": 6866 }, { "epoch": 0.68, "learning_rate": 0.0004992627895850402, "loss": 1.3965, "step": 6867 }, { "epoch": 0.68, "learning_rate": 0.0004989864292449665, "loss": 1.3203, "step": 6868 }, { "epoch": 0.68, "learning_rate": 0.0004987101199844718, "loss": 1.4053, "step": 6869 }, { "epoch": 0.68, "learning_rate": 0.0004984338618317267, "loss": 1.4707, "step": 6870 }, { "epoch": 0.68, "learning_rate": 0.0004981576548148963, "loss": 1.4648, "step": 6871 }, { "epoch": 0.68, "learning_rate": 0.0004978814989621406, "loss": 1.4023, "step": 6872 }, { "epoch": 0.68, "learning_rate": 0.0004976053943016145, "loss": 1.5088, "step": 6873 }, { "epoch": 0.68, "learning_rate": 0.0004973293408614674, "loss": 1.3682, "step": 6874 }, { "epoch": 0.68, "learning_rate": 0.0004970533386698438, "loss": 1.3301, "step": 6875 }, { "epoch": 0.68, "learning_rate": 0.0004967773877548826, "loss": 1.3154, "step": 6876 }, { "epoch": 0.68, "learning_rate": 0.0004965014881447179, "loss": 1.2212, "step": 6877 }, { "epoch": 0.68, "learning_rate": 0.000496225639867478, "loss": 1.4893, "step": 6878 }, { "epoch": 0.68, "learning_rate": 0.0004959498429512864, "loss": 1.5918, "step": 6879 }, { "epoch": 0.68, "learning_rate": 0.0004956740974242618, "loss": 1.6045, "step": 6880 }, { "epoch": 0.68, "learning_rate": 0.0004953984033145169, "loss": 1.4844, "step": 6881 }, { "epoch": 0.68, "learning_rate": 0.0004951227606501589, "loss": 1.4316, "step": 6882 }, { "epoch": 0.68, "learning_rate": 0.000494847169459291, "loss": 1.5166, "step": 6883 }, { "epoch": 0.68, "learning_rate": 0.0004945716297700099, "loss": 1.5439, "step": 6884 }, { "epoch": 0.68, "learning_rate": 0.0004942961416104077, "loss": 1.4229, "step": 6885 }, { "epoch": 0.68, "learning_rate": 0.0004940207050085711, "loss": 1.3496, "step": 6886 }, { "epoch": 0.68, "learning_rate": 0.0004937453199925816, "loss": 1.3184, "step": 6887 }, { "epoch": 0.68, "learning_rate": 0.0004934699865905153, "loss": 1.458, "step": 6888 }, { "epoch": 0.68, "learning_rate": 0.0004931947048304431, "loss": 1.4717, "step": 6889 }, { "epoch": 0.68, "learning_rate": 0.0004929194747404309, "loss": 1.5146, "step": 6890 }, { "epoch": 0.68, "learning_rate": 0.0004926442963485388, "loss": 1.4814, "step": 6891 }, { "epoch": 0.68, "learning_rate": 0.0004923691696828219, "loss": 1.4961, "step": 6892 }, { "epoch": 0.68, "learning_rate": 0.0004920940947713304, "loss": 1.5635, "step": 6893 }, { "epoch": 0.68, "learning_rate": 0.0004918190716421086, "loss": 1.3672, "step": 6894 }, { "epoch": 0.68, "learning_rate": 0.0004915441003231958, "loss": 1.3154, "step": 6895 }, { "epoch": 0.68, "learning_rate": 0.000491269180842626, "loss": 1.4395, "step": 6896 }, { "epoch": 0.68, "learning_rate": 0.0004909943132284278, "loss": 1.4609, "step": 6897 }, { "epoch": 0.68, "learning_rate": 0.0004907194975086251, "loss": 1.5127, "step": 6898 }, { "epoch": 0.68, "learning_rate": 0.0004904447337112357, "loss": 1.5947, "step": 6899 }, { "epoch": 0.68, "learning_rate": 0.0004901700218642725, "loss": 1.2979, "step": 6900 }, { "epoch": 0.68, "learning_rate": 0.0004898953619957431, "loss": 1.4834, "step": 6901 }, { "epoch": 0.68, "learning_rate": 0.0004896207541336498, "loss": 1.498, "step": 6902 }, { "epoch": 0.68, "learning_rate": 0.0004893461983059895, "loss": 1.479, "step": 6903 }, { "epoch": 0.68, "learning_rate": 0.0004890716945407537, "loss": 1.3862, "step": 6904 }, { "epoch": 0.68, "learning_rate": 0.000488797242865929, "loss": 1.4512, "step": 6905 }, { "epoch": 0.68, "learning_rate": 0.0004885228433094962, "loss": 1.4512, "step": 6906 }, { "epoch": 0.68, "learning_rate": 0.0004882484958994311, "loss": 1.4062, "step": 6907 }, { "epoch": 0.68, "learning_rate": 0.00048797420066370425, "loss": 1.5107, "step": 6908 }, { "epoch": 0.68, "learning_rate": 0.00048769995763028054, "loss": 1.3496, "step": 6909 }, { "epoch": 0.68, "learning_rate": 0.0004874257668271197, "loss": 1.5459, "step": 6910 }, { "epoch": 0.68, "learning_rate": 0.0004871516282821764, "loss": 1.3262, "step": 6911 }, { "epoch": 0.68, "learning_rate": 0.0004868775420233995, "loss": 1.521, "step": 6912 }, { "epoch": 0.68, "learning_rate": 0.0004866035080787329, "loss": 1.4521, "step": 6913 }, { "epoch": 0.68, "learning_rate": 0.00048632952647611506, "loss": 1.4316, "step": 6914 }, { "epoch": 0.68, "learning_rate": 0.0004860555972434789, "loss": 1.4844, "step": 6915 }, { "epoch": 0.68, "learning_rate": 0.00048578172040875247, "loss": 1.5176, "step": 6916 }, { "epoch": 0.68, "learning_rate": 0.000485507895999858, "loss": 1.5537, "step": 6917 }, { "epoch": 0.68, "learning_rate": 0.0004852341240447127, "loss": 1.3047, "step": 6918 }, { "epoch": 0.68, "learning_rate": 0.00048496040457122813, "loss": 1.4473, "step": 6919 }, { "epoch": 0.68, "learning_rate": 0.00048468673760731075, "loss": 1.4287, "step": 6920 }, { "epoch": 0.68, "learning_rate": 0.0004844131231808617, "loss": 1.3848, "step": 6921 }, { "epoch": 0.68, "learning_rate": 0.00048413956131977654, "loss": 1.4043, "step": 6922 }, { "epoch": 0.68, "learning_rate": 0.00048386605205194555, "loss": 1.3896, "step": 6923 }, { "epoch": 0.68, "learning_rate": 0.00048359259540525367, "loss": 1.3838, "step": 6924 }, { "epoch": 0.68, "learning_rate": 0.0004833191914075806, "loss": 1.4434, "step": 6925 }, { "epoch": 0.68, "learning_rate": 0.0004830458400868005, "loss": 1.3027, "step": 6926 }, { "epoch": 0.68, "learning_rate": 0.0004827725414707823, "loss": 1.5479, "step": 6927 }, { "epoch": 0.68, "learning_rate": 0.00048249929558738926, "loss": 1.5205, "step": 6928 }, { "epoch": 0.68, "learning_rate": 0.00048222610246447927, "loss": 1.4346, "step": 6929 }, { "epoch": 0.68, "learning_rate": 0.0004819529621299058, "loss": 1.311, "step": 6930 }, { "epoch": 0.68, "learning_rate": 0.00048167987461151564, "loss": 1.3672, "step": 6931 }, { "epoch": 0.68, "learning_rate": 0.00048140683993715095, "loss": 1.5811, "step": 6932 }, { "epoch": 0.68, "learning_rate": 0.00048113385813464826, "loss": 1.4697, "step": 6933 }, { "epoch": 0.68, "learning_rate": 0.00048086092923183867, "loss": 1.5625, "step": 6934 }, { "epoch": 0.68, "learning_rate": 0.000480588053256548, "loss": 1.4639, "step": 6935 }, { "epoch": 0.68, "learning_rate": 0.00048031523023659665, "loss": 1.3086, "step": 6936 }, { "epoch": 0.68, "learning_rate": 0.00048004246019979946, "loss": 1.416, "step": 6937 }, { "epoch": 0.68, "learning_rate": 0.00047976974317396617, "loss": 1.4727, "step": 6938 }, { "epoch": 0.68, "learning_rate": 0.000479497079186901, "loss": 1.4795, "step": 6939 }, { "epoch": 0.68, "learning_rate": 0.00047922446826640253, "loss": 1.4512, "step": 6940 }, { "epoch": 0.68, "learning_rate": 0.00047895191044026433, "loss": 1.5234, "step": 6941 }, { "epoch": 0.68, "learning_rate": 0.0004786794057362742, "loss": 1.3569, "step": 6942 }, { "epoch": 0.68, "learning_rate": 0.0004784069541822147, "loss": 1.4941, "step": 6943 }, { "epoch": 0.68, "learning_rate": 0.00047813455580586304, "loss": 1.3213, "step": 6944 }, { "epoch": 0.68, "learning_rate": 0.0004778622106349908, "loss": 1.4932, "step": 6945 }, { "epoch": 0.68, "learning_rate": 0.0004775899186973642, "loss": 1.4951, "step": 6946 }, { "epoch": 0.68, "learning_rate": 0.0004773176800207439, "loss": 1.4258, "step": 6947 }, { "epoch": 0.68, "learning_rate": 0.0004770454946328859, "loss": 1.4082, "step": 6948 }, { "epoch": 0.69, "learning_rate": 0.0004767733625615399, "loss": 1.4766, "step": 6949 }, { "epoch": 0.69, "learning_rate": 0.00047650128383445035, "loss": 1.3115, "step": 6950 }, { "epoch": 0.69, "learning_rate": 0.0004762292584793564, "loss": 1.4658, "step": 6951 }, { "epoch": 0.69, "learning_rate": 0.0004759572865239917, "loss": 1.4941, "step": 6952 }, { "epoch": 0.69, "learning_rate": 0.0004756853679960844, "loss": 1.3315, "step": 6953 }, { "epoch": 0.69, "learning_rate": 0.0004754135029233575, "loss": 1.4287, "step": 6954 }, { "epoch": 0.69, "learning_rate": 0.000475141691333528, "loss": 1.5303, "step": 6955 }, { "epoch": 0.69, "learning_rate": 0.00047486993325430803, "loss": 1.3975, "step": 6956 }, { "epoch": 0.69, "learning_rate": 0.00047459822871340397, "loss": 1.4629, "step": 6957 }, { "epoch": 0.69, "learning_rate": 0.0004743265777385166, "loss": 1.458, "step": 6958 }, { "epoch": 0.69, "learning_rate": 0.0004740549803573415, "loss": 1.3721, "step": 6959 }, { "epoch": 0.69, "learning_rate": 0.0004737834365975687, "loss": 1.4521, "step": 6960 }, { "epoch": 0.69, "learning_rate": 0.0004735119464868829, "loss": 1.4307, "step": 6961 }, { "epoch": 0.69, "learning_rate": 0.00047324051005296296, "loss": 1.4834, "step": 6962 }, { "epoch": 0.69, "learning_rate": 0.00047296912732348276, "loss": 1.4697, "step": 6963 }, { "epoch": 0.69, "learning_rate": 0.00047269779832611014, "loss": 1.4004, "step": 6964 }, { "epoch": 0.69, "learning_rate": 0.00047242652308850753, "loss": 1.4365, "step": 6965 }, { "epoch": 0.69, "learning_rate": 0.000472155301638333, "loss": 1.4482, "step": 6966 }, { "epoch": 0.69, "learning_rate": 0.0004718841340032376, "loss": 1.4365, "step": 6967 }, { "epoch": 0.69, "learning_rate": 0.00047161302021086815, "loss": 1.4932, "step": 6968 }, { "epoch": 0.69, "learning_rate": 0.00047134196028886455, "loss": 1.5059, "step": 6969 }, { "epoch": 0.69, "learning_rate": 0.00047107095426486247, "loss": 1.4434, "step": 6970 }, { "epoch": 0.69, "learning_rate": 0.00047080000216649155, "loss": 1.4951, "step": 6971 }, { "epoch": 0.69, "learning_rate": 0.00047052910402137606, "loss": 1.3271, "step": 6972 }, { "epoch": 0.69, "learning_rate": 0.00047025825985713465, "loss": 1.4248, "step": 6973 }, { "epoch": 0.69, "learning_rate": 0.0004699874697013807, "loss": 1.5381, "step": 6974 }, { "epoch": 0.69, "learning_rate": 0.00046971673358172197, "loss": 1.4248, "step": 6975 }, { "epoch": 0.69, "learning_rate": 0.0004694460515257606, "loss": 1.415, "step": 6976 }, { "epoch": 0.69, "learning_rate": 0.0004691754235610932, "loss": 1.4932, "step": 6977 }, { "epoch": 0.69, "learning_rate": 0.0004689048497153111, "loss": 1.4355, "step": 6978 }, { "epoch": 0.69, "learning_rate": 0.00046863433001599955, "loss": 1.3936, "step": 6979 }, { "epoch": 0.69, "learning_rate": 0.00046836386449073953, "loss": 1.4512, "step": 6980 }, { "epoch": 0.69, "learning_rate": 0.0004680934531671052, "loss": 1.585, "step": 6981 }, { "epoch": 0.69, "learning_rate": 0.0004678230960726657, "loss": 1.4551, "step": 6982 }, { "epoch": 0.69, "learning_rate": 0.00046755279323498457, "loss": 1.4111, "step": 6983 }, { "epoch": 0.69, "learning_rate": 0.00046728254468162003, "loss": 1.4668, "step": 6984 }, { "epoch": 0.69, "learning_rate": 0.00046701235044012434, "loss": 1.3223, "step": 6985 }, { "epoch": 0.69, "learning_rate": 0.00046674221053804454, "loss": 1.415, "step": 6986 }, { "epoch": 0.69, "learning_rate": 0.00046647212500292224, "loss": 1.3828, "step": 6987 }, { "epoch": 0.69, "learning_rate": 0.0004662020938622932, "loss": 1.3799, "step": 6988 }, { "epoch": 0.69, "learning_rate": 0.00046593211714368777, "loss": 1.4209, "step": 6989 }, { "epoch": 0.69, "learning_rate": 0.0004656621948746307, "loss": 1.457, "step": 6990 }, { "epoch": 0.69, "learning_rate": 0.0004653923270826413, "loss": 1.459, "step": 6991 }, { "epoch": 0.69, "learning_rate": 0.0004651225137952333, "loss": 1.4395, "step": 6992 }, { "epoch": 0.69, "learning_rate": 0.00046485275503991473, "loss": 1.5557, "step": 6993 }, { "epoch": 0.69, "learning_rate": 0.0004645830508441884, "loss": 1.5127, "step": 6994 }, { "epoch": 0.69, "learning_rate": 0.00046431340123555097, "loss": 1.4717, "step": 6995 }, { "epoch": 0.69, "learning_rate": 0.0004640438062414941, "loss": 1.377, "step": 6996 }, { "epoch": 0.69, "learning_rate": 0.0004637742658895032, "loss": 1.4854, "step": 6997 }, { "epoch": 0.69, "learning_rate": 0.00046350478020705955, "loss": 1.5166, "step": 6998 }, { "epoch": 0.69, "learning_rate": 0.0004632353492216372, "loss": 1.4028, "step": 6999 }, { "epoch": 0.69, "learning_rate": 0.0004629659729607055, "loss": 1.3877, "step": 7000 }, { "epoch": 0.69, "learning_rate": 0.00046269665145172814, "loss": 1.5234, "step": 7001 }, { "epoch": 0.69, "learning_rate": 0.0004624273847221627, "loss": 1.4941, "step": 7002 }, { "epoch": 0.69, "learning_rate": 0.000462158172799462, "loss": 1.3857, "step": 7003 }, { "epoch": 0.69, "learning_rate": 0.0004618890157110728, "loss": 1.4219, "step": 7004 }, { "epoch": 0.69, "learning_rate": 0.00046161991348443623, "loss": 1.5049, "step": 7005 }, { "epoch": 0.69, "learning_rate": 0.00046135086614698787, "loss": 1.5068, "step": 7006 }, { "epoch": 0.69, "learning_rate": 0.00046108187372615784, "loss": 1.2764, "step": 7007 }, { "epoch": 0.69, "learning_rate": 0.0004608129362493707, "loss": 1.4307, "step": 7008 }, { "epoch": 0.69, "learning_rate": 0.0004605440537440451, "loss": 1.3633, "step": 7009 }, { "epoch": 0.69, "learning_rate": 0.0004602752262375944, "loss": 1.5234, "step": 7010 }, { "epoch": 0.69, "learning_rate": 0.0004600064537574262, "loss": 1.3662, "step": 7011 }, { "epoch": 0.69, "learning_rate": 0.0004597377363309426, "loss": 1.4902, "step": 7012 }, { "epoch": 0.69, "learning_rate": 0.0004594690739855398, "loss": 1.4146, "step": 7013 }, { "epoch": 0.69, "learning_rate": 0.00045920046674860894, "loss": 1.5225, "step": 7014 }, { "epoch": 0.69, "learning_rate": 0.00045893191464753446, "loss": 1.2695, "step": 7015 }, { "epoch": 0.69, "learning_rate": 0.0004586634177096969, "loss": 1.5098, "step": 7016 }, { "epoch": 0.69, "learning_rate": 0.0004583949759624696, "loss": 1.4062, "step": 7017 }, { "epoch": 0.69, "learning_rate": 0.00045812658943322114, "loss": 1.5625, "step": 7018 }, { "epoch": 0.69, "learning_rate": 0.00045785825814931394, "loss": 1.5361, "step": 7019 }, { "epoch": 0.69, "learning_rate": 0.0004575899821381052, "loss": 1.4043, "step": 7020 }, { "epoch": 0.69, "learning_rate": 0.00045732176142694635, "loss": 1.4062, "step": 7021 }, { "epoch": 0.69, "learning_rate": 0.000457053596043183, "loss": 1.4199, "step": 7022 }, { "epoch": 0.69, "learning_rate": 0.0004567854860141554, "loss": 1.3477, "step": 7023 }, { "epoch": 0.69, "learning_rate": 0.00045651743136719834, "loss": 1.4795, "step": 7024 }, { "epoch": 0.69, "learning_rate": 0.00045624943212964, "loss": 1.5508, "step": 7025 }, { "epoch": 0.69, "learning_rate": 0.0004559814883288038, "loss": 1.5635, "step": 7026 }, { "epoch": 0.69, "learning_rate": 0.00045571359999200736, "loss": 1.3682, "step": 7027 }, { "epoch": 0.69, "learning_rate": 0.0004554457671465626, "loss": 1.3105, "step": 7028 }, { "epoch": 0.69, "learning_rate": 0.00045517798981977543, "loss": 1.457, "step": 7029 }, { "epoch": 0.69, "learning_rate": 0.000454910268038947, "loss": 1.4482, "step": 7030 }, { "epoch": 0.69, "learning_rate": 0.00045464260183137185, "loss": 1.4424, "step": 7031 }, { "epoch": 0.69, "learning_rate": 0.0004543749912243392, "loss": 1.5293, "step": 7032 }, { "epoch": 0.69, "learning_rate": 0.0004541074362451328, "loss": 1.4502, "step": 7033 }, { "epoch": 0.69, "learning_rate": 0.00045383993692103044, "loss": 1.6162, "step": 7034 }, { "epoch": 0.69, "learning_rate": 0.0004535724932793043, "loss": 1.4775, "step": 7035 }, { "epoch": 0.69, "learning_rate": 0.0004533051053472209, "loss": 1.5098, "step": 7036 }, { "epoch": 0.69, "learning_rate": 0.0004530377731520413, "loss": 1.4297, "step": 7037 }, { "epoch": 0.69, "learning_rate": 0.00045277049672102045, "loss": 1.3457, "step": 7038 }, { "epoch": 0.69, "learning_rate": 0.00045250327608140794, "loss": 1.458, "step": 7039 }, { "epoch": 0.69, "learning_rate": 0.0004522361112604476, "loss": 1.3867, "step": 7040 }, { "epoch": 0.69, "learning_rate": 0.00045196900228537754, "loss": 1.5186, "step": 7041 }, { "epoch": 0.69, "learning_rate": 0.0004517019491834301, "loss": 1.542, "step": 7042 }, { "epoch": 0.69, "learning_rate": 0.00045143495198183214, "loss": 1.4121, "step": 7043 }, { "epoch": 0.69, "learning_rate": 0.00045116801070780454, "loss": 1.4717, "step": 7044 }, { "epoch": 0.69, "learning_rate": 0.0004509011253885628, "loss": 1.5498, "step": 7045 }, { "epoch": 0.69, "learning_rate": 0.00045063429605131643, "loss": 1.5059, "step": 7046 }, { "epoch": 0.69, "learning_rate": 0.000450367522723269, "loss": 1.3511, "step": 7047 }, { "epoch": 0.69, "learning_rate": 0.00045010080543161945, "loss": 1.4287, "step": 7048 }, { "epoch": 0.69, "learning_rate": 0.0004498341442035598, "loss": 1.4482, "step": 7049 }, { "epoch": 0.69, "learning_rate": 0.00044956753906627713, "loss": 1.5557, "step": 7050 }, { "epoch": 0.7, "learning_rate": 0.0004493009900469521, "loss": 1.4038, "step": 7051 }, { "epoch": 0.7, "learning_rate": 0.00044903449717276033, "loss": 1.4316, "step": 7052 }, { "epoch": 0.7, "learning_rate": 0.00044876806047087136, "loss": 1.4844, "step": 7053 }, { "epoch": 0.7, "learning_rate": 0.000448501679968449, "loss": 1.4375, "step": 7054 }, { "epoch": 0.7, "learning_rate": 0.0004482353556926515, "loss": 1.3228, "step": 7055 }, { "epoch": 0.7, "learning_rate": 0.0004479690876706313, "loss": 1.3711, "step": 7056 }, { "epoch": 0.7, "learning_rate": 0.000447702875929535, "loss": 1.5205, "step": 7057 }, { "epoch": 0.7, "learning_rate": 0.0004474367204965034, "loss": 1.4609, "step": 7058 }, { "epoch": 0.7, "learning_rate": 0.00044717062139867204, "loss": 1.4229, "step": 7059 }, { "epoch": 0.7, "learning_rate": 0.00044690457866317025, "loss": 1.4834, "step": 7060 }, { "epoch": 0.7, "learning_rate": 0.0004466385923171217, "loss": 1.4932, "step": 7061 }, { "epoch": 0.7, "learning_rate": 0.0004463726623876444, "loss": 1.3447, "step": 7062 }, { "epoch": 0.7, "learning_rate": 0.0004461067889018505, "loss": 1.3594, "step": 7063 }, { "epoch": 0.7, "learning_rate": 0.0004458409718868466, "loss": 1.2646, "step": 7064 }, { "epoch": 0.7, "learning_rate": 0.0004455752113697329, "loss": 1.2583, "step": 7065 }, { "epoch": 0.7, "learning_rate": 0.0004453095073776051, "loss": 1.4727, "step": 7066 }, { "epoch": 0.7, "learning_rate": 0.00044504385993755205, "loss": 1.4873, "step": 7067 }, { "epoch": 0.7, "learning_rate": 0.00044477826907665717, "loss": 1.3896, "step": 7068 }, { "epoch": 0.7, "learning_rate": 0.00044451273482199816, "loss": 1.5537, "step": 7069 }, { "epoch": 0.7, "learning_rate": 0.00044424725720064684, "loss": 1.4307, "step": 7070 }, { "epoch": 0.7, "learning_rate": 0.0004439818362396694, "loss": 1.4629, "step": 7071 }, { "epoch": 0.7, "learning_rate": 0.000443716471966126, "loss": 1.4951, "step": 7072 }, { "epoch": 0.7, "learning_rate": 0.0004434511644070712, "loss": 1.4805, "step": 7073 }, { "epoch": 0.7, "learning_rate": 0.00044318591358955385, "loss": 1.3232, "step": 7074 }, { "epoch": 0.7, "learning_rate": 0.00044292071954061694, "loss": 1.3828, "step": 7075 }, { "epoch": 0.7, "learning_rate": 0.0004426555822872977, "loss": 1.3828, "step": 7076 }, { "epoch": 0.7, "learning_rate": 0.00044239050185662746, "loss": 1.4072, "step": 7077 }, { "epoch": 0.7, "learning_rate": 0.00044212547827563175, "loss": 1.5645, "step": 7078 }, { "epoch": 0.7, "learning_rate": 0.00044186051157133057, "loss": 1.5596, "step": 7079 }, { "epoch": 0.7, "learning_rate": 0.00044159560177073787, "loss": 1.4668, "step": 7080 }, { "epoch": 0.7, "learning_rate": 0.00044133074890086177, "loss": 1.5322, "step": 7081 }, { "epoch": 0.7, "learning_rate": 0.00044106595298870464, "loss": 1.2881, "step": 7082 }, { "epoch": 0.7, "learning_rate": 0.0004408012140612633, "loss": 1.417, "step": 7083 }, { "epoch": 0.7, "learning_rate": 0.0004405365321455285, "loss": 1.3701, "step": 7084 }, { "epoch": 0.7, "learning_rate": 0.000440271907268485, "loss": 1.4658, "step": 7085 }, { "epoch": 0.7, "learning_rate": 0.0004400073394571121, "loss": 1.4922, "step": 7086 }, { "epoch": 0.7, "learning_rate": 0.0004397428287383832, "loss": 1.459, "step": 7087 }, { "epoch": 0.7, "learning_rate": 0.0004394783751392658, "loss": 1.4062, "step": 7088 }, { "epoch": 0.7, "learning_rate": 0.00043921397868672155, "loss": 1.5312, "step": 7089 }, { "epoch": 0.7, "learning_rate": 0.00043894963940770636, "loss": 1.2876, "step": 7090 }, { "epoch": 0.7, "learning_rate": 0.00043868535732917023, "loss": 1.5146, "step": 7091 }, { "epoch": 0.7, "learning_rate": 0.0004384211324780575, "loss": 1.4141, "step": 7092 }, { "epoch": 0.7, "learning_rate": 0.0004381569648813064, "loss": 1.3975, "step": 7093 }, { "epoch": 0.7, "learning_rate": 0.0004378928545658496, "loss": 1.3271, "step": 7094 }, { "epoch": 0.7, "learning_rate": 0.0004376288015586137, "loss": 1.3579, "step": 7095 }, { "epoch": 0.7, "learning_rate": 0.0004373648058865196, "loss": 1.415, "step": 7096 }, { "epoch": 0.7, "learning_rate": 0.0004371008675764824, "loss": 1.4541, "step": 7097 }, { "epoch": 0.7, "learning_rate": 0.0004368369866554107, "loss": 1.5742, "step": 7098 }, { "epoch": 0.7, "learning_rate": 0.0004365731631502088, "loss": 1.376, "step": 7099 }, { "epoch": 0.7, "learning_rate": 0.0004363093970877735, "loss": 1.3623, "step": 7100 }, { "epoch": 0.7, "learning_rate": 0.0004360456884949967, "loss": 1.3955, "step": 7101 }, { "epoch": 0.7, "learning_rate": 0.000435782037398764, "loss": 1.4805, "step": 7102 }, { "epoch": 0.7, "learning_rate": 0.0004355184438259553, "loss": 1.5137, "step": 7103 }, { "epoch": 0.7, "learning_rate": 0.00043525490780344455, "loss": 1.4111, "step": 7104 }, { "epoch": 0.7, "learning_rate": 0.0004349914293581, "loss": 1.46, "step": 7105 }, { "epoch": 0.7, "learning_rate": 0.00043472800851678396, "loss": 1.5352, "step": 7106 }, { "epoch": 0.7, "learning_rate": 0.00043446464530635275, "loss": 1.4863, "step": 7107 }, { "epoch": 0.7, "learning_rate": 0.00043420133975365696, "loss": 1.4238, "step": 7108 }, { "epoch": 0.7, "learning_rate": 0.00043393809188554125, "loss": 1.4814, "step": 7109 }, { "epoch": 0.7, "learning_rate": 0.00043367490172884426, "loss": 1.4863, "step": 7110 }, { "epoch": 0.7, "learning_rate": 0.00043341176931039916, "loss": 1.5068, "step": 7111 }, { "epoch": 0.7, "learning_rate": 0.00043314869465703264, "loss": 1.4512, "step": 7112 }, { "epoch": 0.7, "learning_rate": 0.00043288567779556597, "loss": 1.4482, "step": 7113 }, { "epoch": 0.7, "learning_rate": 0.00043262271875281445, "loss": 1.4688, "step": 7114 }, { "epoch": 0.7, "learning_rate": 0.0004323598175555873, "loss": 1.5869, "step": 7115 }, { "epoch": 0.7, "learning_rate": 0.00043209697423068764, "loss": 1.4473, "step": 7116 }, { "epoch": 0.7, "learning_rate": 0.0004318341888049135, "loss": 1.4277, "step": 7117 }, { "epoch": 0.7, "learning_rate": 0.00043157146130505655, "loss": 1.4775, "step": 7118 }, { "epoch": 0.7, "learning_rate": 0.0004313087917579023, "loss": 1.4355, "step": 7119 }, { "epoch": 0.7, "learning_rate": 0.00043104618019023045, "loss": 1.5244, "step": 7120 }, { "epoch": 0.7, "learning_rate": 0.0004307836266288151, "loss": 1.3135, "step": 7121 }, { "epoch": 0.7, "learning_rate": 0.00043052113110042414, "loss": 1.5596, "step": 7122 }, { "epoch": 0.7, "learning_rate": 0.00043025869363181956, "loss": 1.5186, "step": 7123 }, { "epoch": 0.7, "learning_rate": 0.0004299963142497577, "loss": 1.458, "step": 7124 }, { "epoch": 0.7, "learning_rate": 0.0004297339929809887, "loss": 1.5078, "step": 7125 }, { "epoch": 0.7, "learning_rate": 0.00042947172985225693, "loss": 1.418, "step": 7126 }, { "epoch": 0.7, "learning_rate": 0.0004292095248903006, "loss": 1.3857, "step": 7127 }, { "epoch": 0.7, "learning_rate": 0.00042894737812185235, "loss": 1.3398, "step": 7128 }, { "epoch": 0.7, "learning_rate": 0.0004286852895736385, "loss": 1.5293, "step": 7129 }, { "epoch": 0.7, "learning_rate": 0.0004284232592723798, "loss": 1.4961, "step": 7130 }, { "epoch": 0.7, "learning_rate": 0.00042816128724479087, "loss": 1.4521, "step": 7131 }, { "epoch": 0.7, "learning_rate": 0.00042789937351758023, "loss": 1.3174, "step": 7132 }, { "epoch": 0.7, "learning_rate": 0.0004276375181174509, "loss": 1.4551, "step": 7133 }, { "epoch": 0.7, "learning_rate": 0.0004273757210710991, "loss": 1.4922, "step": 7134 }, { "epoch": 0.7, "learning_rate": 0.0004271139824052166, "loss": 1.5508, "step": 7135 }, { "epoch": 0.7, "learning_rate": 0.00042685230214648787, "loss": 1.4365, "step": 7136 }, { "epoch": 0.7, "learning_rate": 0.0004265906803215922, "loss": 1.417, "step": 7137 }, { "epoch": 0.7, "learning_rate": 0.00042632911695720177, "loss": 1.5488, "step": 7138 }, { "epoch": 0.7, "learning_rate": 0.0004260676120799842, "loss": 1.3799, "step": 7139 }, { "epoch": 0.7, "learning_rate": 0.00042580616571660036, "loss": 1.4961, "step": 7140 }, { "epoch": 0.7, "learning_rate": 0.0004255447778937055, "loss": 1.4648, "step": 7141 }, { "epoch": 0.7, "learning_rate": 0.00042528344863794866, "loss": 1.4219, "step": 7142 }, { "epoch": 0.7, "learning_rate": 0.0004250221779759731, "loss": 1.3789, "step": 7143 }, { "epoch": 0.7, "learning_rate": 0.00042476096593441594, "loss": 1.4072, "step": 7144 }, { "epoch": 0.7, "learning_rate": 0.00042449981253990835, "loss": 1.4482, "step": 7145 }, { "epoch": 0.7, "learning_rate": 0.00042423871781907573, "loss": 1.4043, "step": 7146 }, { "epoch": 0.7, "learning_rate": 0.00042397768179853714, "loss": 1.3701, "step": 7147 }, { "epoch": 0.7, "learning_rate": 0.00042371670450490573, "loss": 1.5381, "step": 7148 }, { "epoch": 0.7, "learning_rate": 0.0004234557859647893, "loss": 1.4766, "step": 7149 }, { "epoch": 0.7, "learning_rate": 0.0004231949262047888, "loss": 1.584, "step": 7150 }, { "epoch": 0.7, "learning_rate": 0.0004229341252514995, "loss": 1.418, "step": 7151 }, { "epoch": 0.71, "learning_rate": 0.00042267338313151104, "loss": 1.4219, "step": 7152 }, { "epoch": 0.71, "learning_rate": 0.00042241269987140627, "loss": 1.4365, "step": 7153 }, { "epoch": 0.71, "learning_rate": 0.0004221520754977628, "loss": 1.4629, "step": 7154 }, { "epoch": 0.71, "learning_rate": 0.00042189151003715176, "loss": 1.2891, "step": 7155 }, { "epoch": 0.71, "learning_rate": 0.0004216310035161386, "loss": 1.4385, "step": 7156 }, { "epoch": 0.71, "learning_rate": 0.0004213705559612825, "loss": 1.312, "step": 7157 }, { "epoch": 0.71, "learning_rate": 0.00042111016739913675, "loss": 1.3486, "step": 7158 }, { "epoch": 0.71, "learning_rate": 0.0004208498378562486, "loss": 1.376, "step": 7159 }, { "epoch": 0.71, "learning_rate": 0.0004205895673591593, "loss": 1.415, "step": 7160 }, { "epoch": 0.71, "learning_rate": 0.00042032935593440423, "loss": 1.4482, "step": 7161 }, { "epoch": 0.71, "learning_rate": 0.0004200692036085123, "loss": 1.4277, "step": 7162 }, { "epoch": 0.71, "learning_rate": 0.00041980911040800695, "loss": 1.5254, "step": 7163 }, { "epoch": 0.71, "learning_rate": 0.0004195490763594052, "loss": 1.4551, "step": 7164 }, { "epoch": 0.71, "learning_rate": 0.00041928910148921806, "loss": 1.377, "step": 7165 }, { "epoch": 0.71, "learning_rate": 0.00041902918582395054, "loss": 1.3271, "step": 7166 }, { "epoch": 0.71, "learning_rate": 0.00041876932939010215, "loss": 1.4004, "step": 7167 }, { "epoch": 0.71, "learning_rate": 0.00041850953221416566, "loss": 1.5195, "step": 7168 }, { "epoch": 0.71, "learning_rate": 0.0004182497943226281, "loss": 1.5684, "step": 7169 }, { "epoch": 0.71, "learning_rate": 0.00041799011574197024, "loss": 1.3818, "step": 7170 }, { "epoch": 0.71, "learning_rate": 0.000417730496498667, "loss": 1.4775, "step": 7171 }, { "epoch": 0.71, "learning_rate": 0.00041747093661918724, "loss": 1.5273, "step": 7172 }, { "epoch": 0.71, "learning_rate": 0.0004172114361299938, "loss": 1.3711, "step": 7173 }, { "epoch": 0.71, "learning_rate": 0.00041695199505754324, "loss": 1.3862, "step": 7174 }, { "epoch": 0.71, "learning_rate": 0.00041669261342828633, "loss": 1.5234, "step": 7175 }, { "epoch": 0.71, "learning_rate": 0.0004164332912686677, "loss": 1.415, "step": 7176 }, { "epoch": 0.71, "learning_rate": 0.00041617402860512587, "loss": 1.4873, "step": 7177 }, { "epoch": 0.71, "learning_rate": 0.0004159148254640932, "loss": 1.498, "step": 7178 }, { "epoch": 0.71, "learning_rate": 0.0004156556818719962, "loss": 1.332, "step": 7179 }, { "epoch": 0.71, "learning_rate": 0.0004153965978552553, "loss": 1.4717, "step": 7180 }, { "epoch": 0.71, "learning_rate": 0.00041513757344028445, "loss": 1.5068, "step": 7181 }, { "epoch": 0.71, "learning_rate": 0.0004148786086534921, "loss": 1.4736, "step": 7182 }, { "epoch": 0.71, "learning_rate": 0.0004146197035212804, "loss": 1.4941, "step": 7183 }, { "epoch": 0.71, "learning_rate": 0.00041436085807004477, "loss": 1.418, "step": 7184 }, { "epoch": 0.71, "learning_rate": 0.00041410207232617604, "loss": 1.415, "step": 7185 }, { "epoch": 0.71, "learning_rate": 0.0004138433463160576, "loss": 1.4082, "step": 7186 }, { "epoch": 0.71, "learning_rate": 0.0004135846800660673, "loss": 1.4541, "step": 7187 }, { "epoch": 0.71, "learning_rate": 0.00041332607360257677, "loss": 1.3184, "step": 7188 }, { "epoch": 0.71, "learning_rate": 0.0004130675269519516, "loss": 1.4795, "step": 7189 }, { "epoch": 0.71, "learning_rate": 0.0004128090401405512, "loss": 1.4492, "step": 7190 }, { "epoch": 0.71, "learning_rate": 0.00041255061319472896, "loss": 1.4331, "step": 7191 }, { "epoch": 0.71, "learning_rate": 0.00041229224614083214, "loss": 1.5859, "step": 7192 }, { "epoch": 0.71, "learning_rate": 0.0004120339390052025, "loss": 1.4561, "step": 7193 }, { "epoch": 0.71, "learning_rate": 0.00041177569181417395, "loss": 1.4424, "step": 7194 }, { "epoch": 0.71, "learning_rate": 0.00041151750459407613, "loss": 1.4043, "step": 7195 }, { "epoch": 0.71, "learning_rate": 0.0004112593773712318, "loss": 1.585, "step": 7196 }, { "epoch": 0.71, "learning_rate": 0.00041100131017195754, "loss": 1.3477, "step": 7197 }, { "epoch": 0.71, "learning_rate": 0.0004107433030225637, "loss": 1.5254, "step": 7198 }, { "epoch": 0.71, "learning_rate": 0.0004104853559493553, "loss": 1.4697, "step": 7199 }, { "epoch": 0.71, "learning_rate": 0.0004102274689786306, "loss": 1.4209, "step": 7200 }, { "epoch": 0.71, "learning_rate": 0.00040996964213668176, "loss": 1.5195, "step": 7201 }, { "epoch": 0.71, "learning_rate": 0.00040971187544979474, "loss": 1.5869, "step": 7202 }, { "epoch": 0.71, "learning_rate": 0.00040945416894424945, "loss": 1.4956, "step": 7203 }, { "epoch": 0.71, "learning_rate": 0.0004091965226463199, "loss": 1.3643, "step": 7204 }, { "epoch": 0.71, "learning_rate": 0.00040893893658227367, "loss": 1.4229, "step": 7205 }, { "epoch": 0.71, "learning_rate": 0.0004086814107783723, "loss": 1.3701, "step": 7206 }, { "epoch": 0.71, "learning_rate": 0.00040842394526087114, "loss": 1.5088, "step": 7207 }, { "epoch": 0.71, "learning_rate": 0.0004081665400560195, "loss": 1.4639, "step": 7208 }, { "epoch": 0.71, "learning_rate": 0.00040790919519006054, "loss": 1.3862, "step": 7209 }, { "epoch": 0.71, "learning_rate": 0.00040765191068923114, "loss": 1.4004, "step": 7210 }, { "epoch": 0.71, "learning_rate": 0.00040739468657976195, "loss": 1.3457, "step": 7211 }, { "epoch": 0.71, "learning_rate": 0.00040713752288787787, "loss": 1.5195, "step": 7212 }, { "epoch": 0.71, "learning_rate": 0.0004068804196397973, "loss": 1.4443, "step": 7213 }, { "epoch": 0.71, "learning_rate": 0.00040662337686173247, "loss": 1.4404, "step": 7214 }, { "epoch": 0.71, "learning_rate": 0.00040636639457988956, "loss": 1.4004, "step": 7215 }, { "epoch": 0.71, "learning_rate": 0.00040610947282046815, "loss": 1.4824, "step": 7216 }, { "epoch": 0.71, "learning_rate": 0.00040585261160966293, "loss": 1.4692, "step": 7217 }, { "epoch": 0.71, "learning_rate": 0.000405595810973661, "loss": 1.4326, "step": 7218 }, { "epoch": 0.71, "learning_rate": 0.00040533907093864396, "loss": 1.4521, "step": 7219 }, { "epoch": 0.71, "learning_rate": 0.0004050823915307871, "loss": 1.3936, "step": 7220 }, { "epoch": 0.71, "learning_rate": 0.0004048257727762593, "loss": 1.3545, "step": 7221 }, { "epoch": 0.71, "learning_rate": 0.0004045692147012238, "loss": 1.3535, "step": 7222 }, { "epoch": 0.71, "learning_rate": 0.00040431271733183704, "loss": 1.543, "step": 7223 }, { "epoch": 0.71, "learning_rate": 0.0004040562806942497, "loss": 1.4844, "step": 7224 }, { "epoch": 0.71, "learning_rate": 0.00040379990481460626, "loss": 1.3115, "step": 7225 }, { "epoch": 0.71, "learning_rate": 0.0004035435897190446, "loss": 1.4219, "step": 7226 }, { "epoch": 0.71, "learning_rate": 0.000403287335433697, "loss": 1.5146, "step": 7227 }, { "epoch": 0.71, "learning_rate": 0.0004030311419846889, "loss": 1.4414, "step": 7228 }, { "epoch": 0.71, "learning_rate": 0.00040277500939814005, "loss": 1.3164, "step": 7229 }, { "epoch": 0.71, "learning_rate": 0.0004025189377001638, "loss": 1.5068, "step": 7230 }, { "epoch": 0.71, "learning_rate": 0.0004022629269168673, "loss": 1.4805, "step": 7231 }, { "epoch": 0.71, "learning_rate": 0.0004020069770743514, "loss": 1.3896, "step": 7232 }, { "epoch": 0.71, "learning_rate": 0.00040175108819871086, "loss": 1.4229, "step": 7233 }, { "epoch": 0.71, "learning_rate": 0.0004014952603160338, "loss": 1.4219, "step": 7234 }, { "epoch": 0.71, "learning_rate": 0.0004012394934524033, "loss": 1.3672, "step": 7235 }, { "epoch": 0.71, "learning_rate": 0.0004009837876338951, "loss": 1.3877, "step": 7236 }, { "epoch": 0.71, "learning_rate": 0.00040072814288657886, "loss": 1.5791, "step": 7237 }, { "epoch": 0.71, "learning_rate": 0.0004004725592365185, "loss": 1.3105, "step": 7238 }, { "epoch": 0.71, "learning_rate": 0.00040021703670977104, "loss": 1.5547, "step": 7239 }, { "epoch": 0.71, "learning_rate": 0.00039996157533238785, "loss": 1.4873, "step": 7240 }, { "epoch": 0.71, "learning_rate": 0.00039970617513041386, "loss": 1.5762, "step": 7241 }, { "epoch": 0.71, "learning_rate": 0.00039945083612988774, "loss": 1.4014, "step": 7242 }, { "epoch": 0.71, "learning_rate": 0.00039919555835684186, "loss": 1.3848, "step": 7243 }, { "epoch": 0.71, "learning_rate": 0.0003989403418373025, "loss": 1.4521, "step": 7244 }, { "epoch": 0.71, "learning_rate": 0.00039868518659728967, "loss": 1.4893, "step": 7245 }, { "epoch": 0.71, "learning_rate": 0.00039843009266281695, "loss": 1.4546, "step": 7246 }, { "epoch": 0.71, "learning_rate": 0.0003981750600598918, "loss": 1.3911, "step": 7247 }, { "epoch": 0.71, "learning_rate": 0.0003979200888145156, "loss": 1.5107, "step": 7248 }, { "epoch": 0.71, "learning_rate": 0.0003976651789526831, "loss": 1.4854, "step": 7249 }, { "epoch": 0.71, "learning_rate": 0.0003974103305003831, "loss": 1.4688, "step": 7250 }, { "epoch": 0.71, "learning_rate": 0.0003971555434835981, "loss": 1.4473, "step": 7251 }, { "epoch": 0.71, "learning_rate": 0.00039690081792830413, "loss": 1.4785, "step": 7252 }, { "epoch": 0.72, "learning_rate": 0.0003966461538604711, "loss": 1.3828, "step": 7253 }, { "epoch": 0.72, "learning_rate": 0.00039639155130606277, "loss": 1.3965, "step": 7254 }, { "epoch": 0.72, "learning_rate": 0.0003961370102910363, "loss": 1.3779, "step": 7255 }, { "epoch": 0.72, "learning_rate": 0.00039588253084134307, "loss": 1.2656, "step": 7256 }, { "epoch": 0.72, "learning_rate": 0.0003956281129829277, "loss": 1.3926, "step": 7257 }, { "epoch": 0.72, "learning_rate": 0.0003953737567417286, "loss": 1.4844, "step": 7258 }, { "epoch": 0.72, "learning_rate": 0.00039511946214367834, "loss": 1.415, "step": 7259 }, { "epoch": 0.72, "learning_rate": 0.0003948652292147027, "loss": 1.3711, "step": 7260 }, { "epoch": 0.72, "learning_rate": 0.00039461105798072125, "loss": 1.4023, "step": 7261 }, { "epoch": 0.72, "learning_rate": 0.0003943569484676477, "loss": 1.5264, "step": 7262 }, { "epoch": 0.72, "learning_rate": 0.00039410290070138876, "loss": 1.4297, "step": 7263 }, { "epoch": 0.72, "learning_rate": 0.0003938489147078456, "loss": 1.4111, "step": 7264 }, { "epoch": 0.72, "learning_rate": 0.00039359499051291245, "loss": 1.3335, "step": 7265 }, { "epoch": 0.72, "learning_rate": 0.0003933411281424772, "loss": 1.4863, "step": 7266 }, { "epoch": 0.72, "learning_rate": 0.0003930873276224227, "loss": 1.457, "step": 7267 }, { "epoch": 0.72, "learning_rate": 0.0003928335889786239, "loss": 1.4971, "step": 7268 }, { "epoch": 0.72, "learning_rate": 0.0003925799122369502, "loss": 1.46, "step": 7269 }, { "epoch": 0.72, "learning_rate": 0.00039232629742326464, "loss": 1.415, "step": 7270 }, { "epoch": 0.72, "learning_rate": 0.0003920727445634237, "loss": 1.3262, "step": 7271 }, { "epoch": 0.72, "learning_rate": 0.00039181925368327785, "loss": 1.3301, "step": 7272 }, { "epoch": 0.72, "learning_rate": 0.0003915658248086712, "loss": 1.4424, "step": 7273 }, { "epoch": 0.72, "learning_rate": 0.00039131245796544123, "loss": 1.4844, "step": 7274 }, { "epoch": 0.72, "learning_rate": 0.00039105915317941953, "loss": 1.459, "step": 7275 }, { "epoch": 0.72, "learning_rate": 0.000390805910476431, "loss": 1.5049, "step": 7276 }, { "epoch": 0.72, "learning_rate": 0.0003905527298822944, "loss": 1.499, "step": 7277 }, { "epoch": 0.72, "learning_rate": 0.00039029961142282224, "loss": 1.5312, "step": 7278 }, { "epoch": 0.72, "learning_rate": 0.00039004655512382046, "loss": 1.3164, "step": 7279 }, { "epoch": 0.72, "learning_rate": 0.00038979356101108875, "loss": 1.3389, "step": 7280 }, { "epoch": 0.72, "learning_rate": 0.0003895406291104207, "loss": 1.4126, "step": 7281 }, { "epoch": 0.72, "learning_rate": 0.000389287759447603, "loss": 1.2588, "step": 7282 }, { "epoch": 0.72, "learning_rate": 0.00038903495204841657, "loss": 1.4502, "step": 7283 }, { "epoch": 0.72, "learning_rate": 0.0003887822069386354, "loss": 1.4404, "step": 7284 }, { "epoch": 0.72, "learning_rate": 0.00038852952414402807, "loss": 1.5586, "step": 7285 }, { "epoch": 0.72, "learning_rate": 0.00038827690369035606, "loss": 1.5903, "step": 7286 }, { "epoch": 0.72, "learning_rate": 0.00038802434560337453, "loss": 1.4189, "step": 7287 }, { "epoch": 0.72, "learning_rate": 0.0003877718499088323, "loss": 1.4561, "step": 7288 }, { "epoch": 0.72, "learning_rate": 0.00038751941663247213, "loss": 1.4219, "step": 7289 }, { "epoch": 0.72, "learning_rate": 0.0003872670458000301, "loss": 1.4521, "step": 7290 }, { "epoch": 0.72, "learning_rate": 0.00038701473743723615, "loss": 1.4717, "step": 7291 }, { "epoch": 0.72, "learning_rate": 0.0003867624915698136, "loss": 1.416, "step": 7292 }, { "epoch": 0.72, "learning_rate": 0.0003865103082234797, "loss": 1.3242, "step": 7293 }, { "epoch": 0.72, "learning_rate": 0.0003862581874239449, "loss": 1.4121, "step": 7294 }, { "epoch": 0.72, "learning_rate": 0.0003860061291969139, "loss": 1.376, "step": 7295 }, { "epoch": 0.72, "learning_rate": 0.0003857541335680844, "loss": 1.2598, "step": 7296 }, { "epoch": 0.72, "learning_rate": 0.0003855022005631481, "loss": 1.4873, "step": 7297 }, { "epoch": 0.72, "learning_rate": 0.0003852503302077902, "loss": 1.3867, "step": 7298 }, { "epoch": 0.72, "learning_rate": 0.0003849985225276895, "loss": 1.5332, "step": 7299 }, { "epoch": 0.72, "learning_rate": 0.0003847467775485184, "loss": 1.5098, "step": 7300 }, { "epoch": 0.72, "learning_rate": 0.00038449509529594285, "loss": 1.5234, "step": 7301 }, { "epoch": 0.72, "learning_rate": 0.00038424347579562223, "loss": 1.5703, "step": 7302 }, { "epoch": 0.72, "learning_rate": 0.0003839919190732105, "loss": 1.4893, "step": 7303 }, { "epoch": 0.72, "learning_rate": 0.00038374042515435403, "loss": 1.5049, "step": 7304 }, { "epoch": 0.72, "learning_rate": 0.0003834889940646936, "loss": 1.3555, "step": 7305 }, { "epoch": 0.72, "learning_rate": 0.0003832376258298628, "loss": 1.4658, "step": 7306 }, { "epoch": 0.72, "learning_rate": 0.0003829863204754892, "loss": 1.374, "step": 7307 }, { "epoch": 0.72, "learning_rate": 0.00038273507802719434, "loss": 1.4624, "step": 7308 }, { "epoch": 0.72, "learning_rate": 0.00038248389851059273, "loss": 1.4688, "step": 7309 }, { "epoch": 0.72, "learning_rate": 0.000382232781951293, "loss": 1.5186, "step": 7310 }, { "epoch": 0.72, "learning_rate": 0.00038198172837489684, "loss": 1.3994, "step": 7311 }, { "epoch": 0.72, "learning_rate": 0.00038173073780700006, "loss": 1.4453, "step": 7312 }, { "epoch": 0.72, "learning_rate": 0.0003814798102731917, "loss": 1.291, "step": 7313 }, { "epoch": 0.72, "learning_rate": 0.0003812289457990543, "loss": 1.5566, "step": 7314 }, { "epoch": 0.72, "learning_rate": 0.0003809781444101643, "loss": 1.3496, "step": 7315 }, { "epoch": 0.72, "learning_rate": 0.0003807274061320915, "loss": 1.5469, "step": 7316 }, { "epoch": 0.72, "learning_rate": 0.0003804767309903987, "loss": 1.4707, "step": 7317 }, { "epoch": 0.72, "learning_rate": 0.00038022611901064384, "loss": 1.3984, "step": 7318 }, { "epoch": 0.72, "learning_rate": 0.0003799755702183769, "loss": 1.4912, "step": 7319 }, { "epoch": 0.72, "learning_rate": 0.00037972508463914213, "loss": 1.3965, "step": 7320 }, { "epoch": 0.72, "learning_rate": 0.00037947466229847705, "loss": 1.4111, "step": 7321 }, { "epoch": 0.72, "learning_rate": 0.0003792243032219128, "loss": 1.5342, "step": 7322 }, { "epoch": 0.72, "learning_rate": 0.0003789740074349741, "loss": 1.3857, "step": 7323 }, { "epoch": 0.72, "learning_rate": 0.00037872377496317943, "loss": 1.457, "step": 7324 }, { "epoch": 0.72, "learning_rate": 0.0003784736058320404, "loss": 1.4854, "step": 7325 }, { "epoch": 0.72, "learning_rate": 0.0003782235000670624, "loss": 1.5674, "step": 7326 }, { "epoch": 0.72, "learning_rate": 0.0003779734576937445, "loss": 1.3242, "step": 7327 }, { "epoch": 0.72, "learning_rate": 0.00037772347873757893, "loss": 1.4072, "step": 7328 }, { "epoch": 0.72, "learning_rate": 0.0003774735632240518, "loss": 1.4355, "step": 7329 }, { "epoch": 0.72, "learning_rate": 0.0003772237111786425, "loss": 1.373, "step": 7330 }, { "epoch": 0.72, "learning_rate": 0.0003769739226268243, "loss": 1.4336, "step": 7331 }, { "epoch": 0.72, "learning_rate": 0.0003767241975940635, "loss": 1.334, "step": 7332 }, { "epoch": 0.72, "learning_rate": 0.0003764745361058203, "loss": 1.457, "step": 7333 }, { "epoch": 0.72, "learning_rate": 0.0003762249381875482, "loss": 1.416, "step": 7334 }, { "epoch": 0.72, "learning_rate": 0.0003759754038646943, "loss": 1.4229, "step": 7335 }, { "epoch": 0.72, "learning_rate": 0.00037572593316269955, "loss": 1.5684, "step": 7336 }, { "epoch": 0.72, "learning_rate": 0.000375476526106998, "loss": 1.5186, "step": 7337 }, { "epoch": 0.72, "learning_rate": 0.0003752271827230173, "loss": 1.3965, "step": 7338 }, { "epoch": 0.72, "learning_rate": 0.0003749779030361785, "loss": 1.293, "step": 7339 }, { "epoch": 0.72, "learning_rate": 0.00037472868707189645, "loss": 1.4531, "step": 7340 }, { "epoch": 0.72, "learning_rate": 0.0003744795348555792, "loss": 1.249, "step": 7341 }, { "epoch": 0.72, "learning_rate": 0.0003742304464126285, "loss": 1.5186, "step": 7342 }, { "epoch": 0.72, "learning_rate": 0.0003739814217684395, "loss": 1.4443, "step": 7343 }, { "epoch": 0.72, "learning_rate": 0.00037373246094840097, "loss": 1.458, "step": 7344 }, { "epoch": 0.72, "learning_rate": 0.00037348356397789496, "loss": 1.3252, "step": 7345 }, { "epoch": 0.72, "learning_rate": 0.0003732347308822972, "loss": 1.5557, "step": 7346 }, { "epoch": 0.72, "learning_rate": 0.00037298596168697695, "loss": 1.4355, "step": 7347 }, { "epoch": 0.72, "learning_rate": 0.0003727372564172966, "loss": 1.3955, "step": 7348 }, { "epoch": 0.72, "learning_rate": 0.00037248861509861254, "loss": 1.4658, "step": 7349 }, { "epoch": 0.72, "learning_rate": 0.00037224003775627415, "loss": 1.3857, "step": 7350 }, { "epoch": 0.72, "learning_rate": 0.00037199152441562454, "loss": 1.502, "step": 7351 }, { "epoch": 0.72, "learning_rate": 0.0003717430751020005, "loss": 1.3799, "step": 7352 }, { "epoch": 0.72, "learning_rate": 0.0003714946898407314, "loss": 1.4414, "step": 7353 }, { "epoch": 0.72, "learning_rate": 0.0003712463686571416, "loss": 1.415, "step": 7354 }, { "epoch": 0.73, "learning_rate": 0.00037099811157654763, "loss": 1.2412, "step": 7355 }, { "epoch": 0.73, "learning_rate": 0.00037074991862426, "loss": 1.5, "step": 7356 }, { "epoch": 0.73, "learning_rate": 0.00037050178982558256, "loss": 1.4736, "step": 7357 }, { "epoch": 0.73, "learning_rate": 0.0003702537252058126, "loss": 1.3955, "step": 7358 }, { "epoch": 0.73, "learning_rate": 0.0003700057247902411, "loss": 1.188, "step": 7359 }, { "epoch": 0.73, "learning_rate": 0.0003697577886041521, "loss": 1.5166, "step": 7360 }, { "epoch": 0.73, "learning_rate": 0.00036950991667282376, "loss": 1.3311, "step": 7361 }, { "epoch": 0.73, "learning_rate": 0.0003692621090215266, "loss": 1.3848, "step": 7362 }, { "epoch": 0.73, "learning_rate": 0.0003690143656755255, "loss": 1.375, "step": 7363 }, { "epoch": 0.73, "learning_rate": 0.0003687666866600785, "loss": 1.5889, "step": 7364 }, { "epoch": 0.73, "learning_rate": 0.0003685190720004371, "loss": 1.4316, "step": 7365 }, { "epoch": 0.73, "learning_rate": 0.0003682715217218463, "loss": 1.5713, "step": 7366 }, { "epoch": 0.73, "learning_rate": 0.00036802403584954404, "loss": 1.4199, "step": 7367 }, { "epoch": 0.73, "learning_rate": 0.00036777661440876274, "loss": 1.3643, "step": 7368 }, { "epoch": 0.73, "learning_rate": 0.00036752925742472754, "loss": 1.4575, "step": 7369 }, { "epoch": 0.73, "learning_rate": 0.00036728196492265687, "loss": 1.4268, "step": 7370 }, { "epoch": 0.73, "learning_rate": 0.0003670347369277629, "loss": 1.4805, "step": 7371 }, { "epoch": 0.73, "learning_rate": 0.0003667875734652512, "loss": 1.4795, "step": 7372 }, { "epoch": 0.73, "learning_rate": 0.00036654047456032056, "loss": 1.4961, "step": 7373 }, { "epoch": 0.73, "learning_rate": 0.00036629344023816337, "loss": 1.375, "step": 7374 }, { "epoch": 0.73, "learning_rate": 0.00036604647052396546, "loss": 1.3701, "step": 7375 }, { "epoch": 0.73, "learning_rate": 0.000365799565442906, "loss": 1.4434, "step": 7376 }, { "epoch": 0.73, "learning_rate": 0.00036555272502015746, "loss": 1.4668, "step": 7377 }, { "epoch": 0.73, "learning_rate": 0.000365305949280886, "loss": 1.4648, "step": 7378 }, { "epoch": 0.73, "learning_rate": 0.0003650592382502509, "loss": 1.4365, "step": 7379 }, { "epoch": 0.73, "learning_rate": 0.00036481259195340497, "loss": 1.5557, "step": 7380 }, { "epoch": 0.73, "learning_rate": 0.0003645660104154945, "loss": 1.4434, "step": 7381 }, { "epoch": 0.73, "learning_rate": 0.00036431949366165896, "loss": 1.5615, "step": 7382 }, { "epoch": 0.73, "learning_rate": 0.0003640730417170315, "loss": 1.4854, "step": 7383 }, { "epoch": 0.73, "learning_rate": 0.00036382665460673837, "loss": 1.4199, "step": 7384 }, { "epoch": 0.73, "learning_rate": 0.00036358033235589916, "loss": 1.4531, "step": 7385 }, { "epoch": 0.73, "learning_rate": 0.00036333407498962756, "loss": 1.4668, "step": 7386 }, { "epoch": 0.73, "learning_rate": 0.00036308788253302984, "loss": 1.417, "step": 7387 }, { "epoch": 0.73, "learning_rate": 0.0003628417550112059, "loss": 1.4609, "step": 7388 }, { "epoch": 0.73, "learning_rate": 0.00036259569244924915, "loss": 1.3896, "step": 7389 }, { "epoch": 0.73, "learning_rate": 0.0003623496948722461, "loss": 1.4248, "step": 7390 }, { "epoch": 0.73, "learning_rate": 0.0003621037623052771, "loss": 1.415, "step": 7391 }, { "epoch": 0.73, "learning_rate": 0.00036185789477341534, "loss": 1.4023, "step": 7392 }, { "epoch": 0.73, "learning_rate": 0.00036161209230172776, "loss": 1.4971, "step": 7393 }, { "epoch": 0.73, "learning_rate": 0.00036136635491527437, "loss": 1.4316, "step": 7394 }, { "epoch": 0.73, "learning_rate": 0.0003611206826391089, "loss": 1.54, "step": 7395 }, { "epoch": 0.73, "learning_rate": 0.0003608750754982781, "loss": 1.4033, "step": 7396 }, { "epoch": 0.73, "learning_rate": 0.0003606295335178224, "loss": 1.4473, "step": 7397 }, { "epoch": 0.73, "learning_rate": 0.0003603840567227753, "loss": 1.4541, "step": 7398 }, { "epoch": 0.73, "learning_rate": 0.00036013864513816373, "loss": 1.4844, "step": 7399 }, { "epoch": 0.73, "learning_rate": 0.000359893298789008, "loss": 1.5137, "step": 7400 }, { "epoch": 0.73, "learning_rate": 0.00035964801770032196, "loss": 1.5459, "step": 7401 }, { "epoch": 0.73, "learning_rate": 0.0003594028018971124, "loss": 1.5596, "step": 7402 }, { "epoch": 0.73, "learning_rate": 0.0003591576514043795, "loss": 1.373, "step": 7403 }, { "epoch": 0.73, "learning_rate": 0.00035891256624711763, "loss": 1.3384, "step": 7404 }, { "epoch": 0.73, "learning_rate": 0.0003586675464503134, "loss": 1.4756, "step": 7405 }, { "epoch": 0.73, "learning_rate": 0.0003584225920389472, "loss": 1.3477, "step": 7406 }, { "epoch": 0.73, "learning_rate": 0.0003581777030379929, "loss": 1.4102, "step": 7407 }, { "epoch": 0.73, "learning_rate": 0.00035793287947241735, "loss": 1.3584, "step": 7408 }, { "epoch": 0.73, "learning_rate": 0.0003576881213671811, "loss": 1.5264, "step": 7409 }, { "epoch": 0.73, "learning_rate": 0.00035744342874723757, "loss": 1.4375, "step": 7410 }, { "epoch": 0.73, "learning_rate": 0.00035719880163753405, "loss": 1.4141, "step": 7411 }, { "epoch": 0.73, "learning_rate": 0.0003569542400630108, "loss": 1.4219, "step": 7412 }, { "epoch": 0.73, "learning_rate": 0.00035670974404860146, "loss": 1.3174, "step": 7413 }, { "epoch": 0.73, "learning_rate": 0.00035646531361923294, "loss": 1.5723, "step": 7414 }, { "epoch": 0.73, "learning_rate": 0.0003562209487998257, "loss": 1.4482, "step": 7415 }, { "epoch": 0.73, "learning_rate": 0.00035597664961529307, "loss": 1.4492, "step": 7416 }, { "epoch": 0.73, "learning_rate": 0.00035573241609054227, "loss": 1.374, "step": 7417 }, { "epoch": 0.73, "learning_rate": 0.00035548824825047345, "loss": 1.5723, "step": 7418 }, { "epoch": 0.73, "learning_rate": 0.0003552441461199799, "loss": 1.4512, "step": 7419 }, { "epoch": 0.73, "learning_rate": 0.0003550001097239487, "loss": 1.4316, "step": 7420 }, { "epoch": 0.73, "learning_rate": 0.0003547561390872598, "loss": 1.4805, "step": 7421 }, { "epoch": 0.73, "learning_rate": 0.00035451223423478676, "loss": 1.4365, "step": 7422 }, { "epoch": 0.73, "learning_rate": 0.0003542683951913961, "loss": 1.2251, "step": 7423 }, { "epoch": 0.73, "learning_rate": 0.00035402462198194797, "loss": 1.3643, "step": 7424 }, { "epoch": 0.73, "learning_rate": 0.0003537809146312956, "loss": 1.3848, "step": 7425 }, { "epoch": 0.73, "learning_rate": 0.00035353727316428553, "loss": 1.5049, "step": 7426 }, { "epoch": 0.73, "learning_rate": 0.0003532936976057576, "loss": 1.3613, "step": 7427 }, { "epoch": 0.73, "learning_rate": 0.00035305018798054513, "loss": 1.4443, "step": 7428 }, { "epoch": 0.73, "learning_rate": 0.00035280674431347424, "loss": 1.498, "step": 7429 }, { "epoch": 0.73, "learning_rate": 0.00035256336662936485, "loss": 1.4336, "step": 7430 }, { "epoch": 0.73, "learning_rate": 0.00035232005495302985, "loss": 1.334, "step": 7431 }, { "epoch": 0.73, "learning_rate": 0.00035207680930927546, "loss": 1.4111, "step": 7432 }, { "epoch": 0.73, "learning_rate": 0.0003518336297229011, "loss": 1.5156, "step": 7433 }, { "epoch": 0.73, "learning_rate": 0.00035159051621869977, "loss": 1.4541, "step": 7434 }, { "epoch": 0.73, "learning_rate": 0.00035134746882145687, "loss": 1.5068, "step": 7435 }, { "epoch": 0.73, "learning_rate": 0.00035110448755595246, "loss": 1.4521, "step": 7436 }, { "epoch": 0.73, "learning_rate": 0.0003508615724469588, "loss": 1.4512, "step": 7437 }, { "epoch": 0.73, "learning_rate": 0.0003506187235192417, "loss": 1.4014, "step": 7438 }, { "epoch": 0.73, "learning_rate": 0.0003503759407975602, "loss": 1.334, "step": 7439 }, { "epoch": 0.73, "learning_rate": 0.0003501332243066666, "loss": 1.4092, "step": 7440 }, { "epoch": 0.73, "learning_rate": 0.00034989057407130644, "loss": 1.3374, "step": 7441 }, { "epoch": 0.73, "learning_rate": 0.0003496479901162184, "loss": 1.4521, "step": 7442 }, { "epoch": 0.73, "learning_rate": 0.00034940547246613464, "loss": 1.4844, "step": 7443 }, { "epoch": 0.73, "learning_rate": 0.0003491630211457805, "loss": 1.4766, "step": 7444 }, { "epoch": 0.73, "learning_rate": 0.00034892063617987436, "loss": 1.6123, "step": 7445 }, { "epoch": 0.73, "learning_rate": 0.00034867831759312796, "loss": 1.3848, "step": 7446 }, { "epoch": 0.73, "learning_rate": 0.00034843606541024643, "loss": 1.4775, "step": 7447 }, { "epoch": 0.73, "learning_rate": 0.00034819387965592797, "loss": 1.4727, "step": 7448 }, { "epoch": 0.73, "learning_rate": 0.00034795176035486386, "loss": 1.4404, "step": 7449 }, { "epoch": 0.73, "learning_rate": 0.0003477097075317388, "loss": 1.332, "step": 7450 }, { "epoch": 0.73, "learning_rate": 0.00034746772121123083, "loss": 1.3809, "step": 7451 }, { "epoch": 0.73, "learning_rate": 0.00034722580141801084, "loss": 1.3945, "step": 7452 }, { "epoch": 0.73, "learning_rate": 0.00034698394817674305, "loss": 1.4688, "step": 7453 }, { "epoch": 0.73, "learning_rate": 0.0003467421615120855, "loss": 1.4512, "step": 7454 }, { "epoch": 0.73, "learning_rate": 0.00034650044144868866, "loss": 1.4863, "step": 7455 }, { "epoch": 0.74, "learning_rate": 0.0003462587880111966, "loss": 1.4268, "step": 7456 }, { "epoch": 0.74, "learning_rate": 0.0003460172012242463, "loss": 1.4092, "step": 7457 }, { "epoch": 0.74, "learning_rate": 0.00034577568111246817, "loss": 1.4189, "step": 7458 }, { "epoch": 0.74, "learning_rate": 0.0003455342277004858, "loss": 1.4141, "step": 7459 }, { "epoch": 0.74, "learning_rate": 0.00034529284101291613, "loss": 1.5312, "step": 7460 }, { "epoch": 0.74, "learning_rate": 0.0003450515210743689, "loss": 1.3828, "step": 7461 }, { "epoch": 0.74, "learning_rate": 0.0003448102679094475, "loss": 1.3418, "step": 7462 }, { "epoch": 0.74, "learning_rate": 0.000344569081542748, "loss": 1.2734, "step": 7463 }, { "epoch": 0.74, "learning_rate": 0.0003443279619988602, "loss": 1.3369, "step": 7464 }, { "epoch": 0.74, "learning_rate": 0.00034408690930236676, "loss": 1.4639, "step": 7465 }, { "epoch": 0.74, "learning_rate": 0.0003438459234778436, "loss": 1.3984, "step": 7466 }, { "epoch": 0.74, "learning_rate": 0.00034360500454985987, "loss": 1.4277, "step": 7467 }, { "epoch": 0.74, "learning_rate": 0.0003433641525429777, "loss": 1.4062, "step": 7468 }, { "epoch": 0.74, "learning_rate": 0.0003431233674817528, "loss": 1.377, "step": 7469 }, { "epoch": 0.74, "learning_rate": 0.00034288264939073344, "loss": 1.4727, "step": 7470 }, { "epoch": 0.74, "learning_rate": 0.0003426419982944615, "loss": 1.4707, "step": 7471 }, { "epoch": 0.74, "learning_rate": 0.0003424014142174724, "loss": 1.3926, "step": 7472 }, { "epoch": 0.74, "learning_rate": 0.00034216089718429386, "loss": 1.3652, "step": 7473 }, { "epoch": 0.74, "learning_rate": 0.00034192044721944784, "loss": 1.457, "step": 7474 }, { "epoch": 0.74, "learning_rate": 0.0003416800643474479, "loss": 1.3916, "step": 7475 }, { "epoch": 0.74, "learning_rate": 0.000341439748592802, "loss": 1.457, "step": 7476 }, { "epoch": 0.74, "learning_rate": 0.0003411994999800111, "loss": 1.5, "step": 7477 }, { "epoch": 0.74, "learning_rate": 0.0003409593185335689, "loss": 1.4287, "step": 7478 }, { "epoch": 0.74, "learning_rate": 0.0003407192042779628, "loss": 1.4219, "step": 7479 }, { "epoch": 0.74, "learning_rate": 0.0003404791572376726, "loss": 1.3125, "step": 7480 }, { "epoch": 0.74, "learning_rate": 0.00034023917743717226, "loss": 1.3525, "step": 7481 }, { "epoch": 0.74, "learning_rate": 0.0003399992649009279, "loss": 1.4424, "step": 7482 }, { "epoch": 0.74, "learning_rate": 0.0003397594196533993, "loss": 1.3643, "step": 7483 }, { "epoch": 0.74, "learning_rate": 0.0003395196417190394, "loss": 1.3604, "step": 7484 }, { "epoch": 0.74, "learning_rate": 0.0003392799311222936, "loss": 1.3936, "step": 7485 }, { "epoch": 0.74, "learning_rate": 0.00033904028788760187, "loss": 1.5049, "step": 7486 }, { "epoch": 0.74, "learning_rate": 0.000338800712039396, "loss": 1.4199, "step": 7487 }, { "epoch": 0.74, "learning_rate": 0.0003385612036021014, "loss": 1.3633, "step": 7488 }, { "epoch": 0.74, "learning_rate": 0.00033832176260013646, "loss": 1.4395, "step": 7489 }, { "epoch": 0.74, "learning_rate": 0.0003380823890579129, "loss": 1.3608, "step": 7490 }, { "epoch": 0.74, "learning_rate": 0.00033784308299983525, "loss": 1.4785, "step": 7491 }, { "epoch": 0.74, "learning_rate": 0.0003376038444503016, "loss": 1.3823, "step": 7492 }, { "epoch": 0.74, "learning_rate": 0.0003373646734337028, "loss": 1.4561, "step": 7493 }, { "epoch": 0.74, "learning_rate": 0.0003371255699744229, "loss": 1.354, "step": 7494 }, { "epoch": 0.74, "learning_rate": 0.00033688653409683913, "loss": 1.4395, "step": 7495 }, { "epoch": 0.74, "learning_rate": 0.00033664756582532174, "loss": 1.4131, "step": 7496 }, { "epoch": 0.74, "learning_rate": 0.0003364086651842343, "loss": 1.3418, "step": 7497 }, { "epoch": 0.74, "learning_rate": 0.00033616983219793316, "loss": 1.4814, "step": 7498 }, { "epoch": 0.74, "learning_rate": 0.000335931066890768, "loss": 1.3955, "step": 7499 }, { "epoch": 0.74, "learning_rate": 0.00033569236928708145, "loss": 1.4346, "step": 7500 }, { "epoch": 0.74, "learning_rate": 0.0003354537394112095, "loss": 1.3867, "step": 7501 }, { "epoch": 0.74, "learning_rate": 0.00033521517728748096, "loss": 1.4443, "step": 7502 }, { "epoch": 0.74, "learning_rate": 0.0003349766829402174, "loss": 1.5254, "step": 7503 }, { "epoch": 0.74, "learning_rate": 0.0003347382563937348, "loss": 1.4395, "step": 7504 }, { "epoch": 0.74, "learning_rate": 0.0003344998976723408, "loss": 1.5684, "step": 7505 }, { "epoch": 0.74, "learning_rate": 0.0003342616068003368, "loss": 1.5371, "step": 7506 }, { "epoch": 0.74, "learning_rate": 0.00033402338380201705, "loss": 1.4932, "step": 7507 }, { "epoch": 0.74, "learning_rate": 0.00033378522870166906, "loss": 1.4082, "step": 7508 }, { "epoch": 0.74, "learning_rate": 0.00033354714152357325, "loss": 1.3301, "step": 7509 }, { "epoch": 0.74, "learning_rate": 0.0003333091222920033, "loss": 1.2734, "step": 7510 }, { "epoch": 0.74, "learning_rate": 0.00033307117103122575, "loss": 1.3027, "step": 7511 }, { "epoch": 0.74, "learning_rate": 0.0003328332877655004, "loss": 1.2812, "step": 7512 }, { "epoch": 0.74, "learning_rate": 0.00033259547251907994, "loss": 1.4932, "step": 7513 }, { "epoch": 0.74, "learning_rate": 0.00033235772531621044, "loss": 1.3779, "step": 7514 }, { "epoch": 0.74, "learning_rate": 0.0003321200461811306, "loss": 1.3477, "step": 7515 }, { "epoch": 0.74, "learning_rate": 0.00033188243513807247, "loss": 1.3311, "step": 7516 }, { "epoch": 0.74, "learning_rate": 0.00033164489221126117, "loss": 1.418, "step": 7517 }, { "epoch": 0.74, "learning_rate": 0.00033140741742491476, "loss": 1.4268, "step": 7518 }, { "epoch": 0.74, "learning_rate": 0.00033117001080324425, "loss": 1.6035, "step": 7519 }, { "epoch": 0.74, "learning_rate": 0.000330932672370454, "loss": 1.4102, "step": 7520 }, { "epoch": 0.74, "learning_rate": 0.0003306954021507408, "loss": 1.4893, "step": 7521 }, { "epoch": 0.74, "learning_rate": 0.0003304582001682956, "loss": 1.4189, "step": 7522 }, { "epoch": 0.74, "learning_rate": 0.00033022106644730155, "loss": 1.3779, "step": 7523 }, { "epoch": 0.74, "learning_rate": 0.0003299840010119349, "loss": 1.4424, "step": 7524 }, { "epoch": 0.74, "learning_rate": 0.00032974700388636505, "loss": 1.3027, "step": 7525 }, { "epoch": 0.74, "learning_rate": 0.0003295100750947545, "loss": 1.4375, "step": 7526 }, { "epoch": 0.74, "learning_rate": 0.0003292732146612588, "loss": 1.3906, "step": 7527 }, { "epoch": 0.74, "learning_rate": 0.0003290364226100263, "loss": 1.4873, "step": 7528 }, { "epoch": 0.74, "learning_rate": 0.00032879969896519855, "loss": 1.394, "step": 7529 }, { "epoch": 0.74, "learning_rate": 0.00032856304375091064, "loss": 1.4268, "step": 7530 }, { "epoch": 0.74, "learning_rate": 0.00032832645699128925, "loss": 1.4121, "step": 7531 }, { "epoch": 0.74, "learning_rate": 0.00032808993871045555, "loss": 1.3594, "step": 7532 }, { "epoch": 0.74, "learning_rate": 0.00032785348893252296, "loss": 1.2319, "step": 7533 }, { "epoch": 0.74, "learning_rate": 0.0003276171076815982, "loss": 1.4316, "step": 7534 }, { "epoch": 0.74, "learning_rate": 0.000327380794981781, "loss": 1.3389, "step": 7535 }, { "epoch": 0.74, "learning_rate": 0.00032714455085716356, "loss": 1.4678, "step": 7536 }, { "epoch": 0.74, "learning_rate": 0.0003269083753318324, "loss": 1.4141, "step": 7537 }, { "epoch": 0.74, "learning_rate": 0.00032667226842986565, "loss": 1.4014, "step": 7538 }, { "epoch": 0.74, "learning_rate": 0.0003264362301753352, "loss": 1.4014, "step": 7539 }, { "epoch": 0.74, "learning_rate": 0.00032620026059230556, "loss": 1.4053, "step": 7540 }, { "epoch": 0.74, "learning_rate": 0.00032596435970483453, "loss": 1.498, "step": 7541 }, { "epoch": 0.74, "learning_rate": 0.0003257285275369728, "loss": 1.5137, "step": 7542 }, { "epoch": 0.74, "learning_rate": 0.000325492764112764, "loss": 1.4424, "step": 7543 }, { "epoch": 0.74, "learning_rate": 0.00032525706945624476, "loss": 1.4365, "step": 7544 }, { "epoch": 0.74, "learning_rate": 0.0003250214435914448, "loss": 1.5371, "step": 7545 }, { "epoch": 0.74, "learning_rate": 0.0003247858865423867, "loss": 1.5322, "step": 7546 }, { "epoch": 0.74, "learning_rate": 0.0003245503983330863, "loss": 1.1982, "step": 7547 }, { "epoch": 0.74, "learning_rate": 0.0003243149789875519, "loss": 1.3076, "step": 7548 }, { "epoch": 0.74, "learning_rate": 0.00032407962852978525, "loss": 1.4512, "step": 7549 }, { "epoch": 0.74, "learning_rate": 0.00032384434698378084, "loss": 1.4033, "step": 7550 }, { "epoch": 0.74, "learning_rate": 0.00032360913437352634, "loss": 1.4102, "step": 7551 }, { "epoch": 0.74, "learning_rate": 0.0003233739907230021, "loss": 1.4697, "step": 7552 }, { "epoch": 0.74, "learning_rate": 0.00032313891605618186, "loss": 1.3457, "step": 7553 }, { "epoch": 0.74, "learning_rate": 0.0003229039103970314, "loss": 1.3926, "step": 7554 }, { "epoch": 0.74, "learning_rate": 0.000322668973769511, "loss": 1.3701, "step": 7555 }, { "epoch": 0.74, "learning_rate": 0.0003224341061975726, "loss": 1.5801, "step": 7556 }, { "epoch": 0.74, "learning_rate": 0.0003221993077051616, "loss": 1.4209, "step": 7557 }, { "epoch": 0.75, "learning_rate": 0.0003219645783162163, "loss": 1.3086, "step": 7558 }, { "epoch": 0.75, "learning_rate": 0.0003217299180546679, "loss": 1.374, "step": 7559 }, { "epoch": 0.75, "learning_rate": 0.00032149532694444063, "loss": 1.3848, "step": 7560 }, { "epoch": 0.75, "learning_rate": 0.00032126080500945165, "loss": 1.3574, "step": 7561 }, { "epoch": 0.75, "learning_rate": 0.00032102635227361097, "loss": 1.5312, "step": 7562 }, { "epoch": 0.75, "learning_rate": 0.00032079196876082174, "loss": 1.5459, "step": 7563 }, { "epoch": 0.75, "learning_rate": 0.0003205576544949799, "loss": 1.3164, "step": 7564 }, { "epoch": 0.75, "learning_rate": 0.0003203234094999742, "loss": 1.4297, "step": 7565 }, { "epoch": 0.75, "learning_rate": 0.0003200892337996868, "loss": 1.4087, "step": 7566 }, { "epoch": 0.75, "learning_rate": 0.0003198551274179924, "loss": 1.3184, "step": 7567 }, { "epoch": 0.75, "learning_rate": 0.00031962109037875854, "loss": 1.4102, "step": 7568 }, { "epoch": 0.75, "learning_rate": 0.00031938712270584616, "loss": 1.627, "step": 7569 }, { "epoch": 0.75, "learning_rate": 0.0003191532244231086, "loss": 1.4512, "step": 7570 }, { "epoch": 0.75, "learning_rate": 0.00031891939555439254, "loss": 1.5586, "step": 7571 }, { "epoch": 0.75, "learning_rate": 0.00031868563612353707, "loss": 1.4316, "step": 7572 }, { "epoch": 0.75, "learning_rate": 0.000318451946154375, "loss": 1.4453, "step": 7573 }, { "epoch": 0.75, "learning_rate": 0.0003182183256707316, "loss": 1.4912, "step": 7574 }, { "epoch": 0.75, "learning_rate": 0.00031798477469642474, "loss": 1.4434, "step": 7575 }, { "epoch": 0.75, "learning_rate": 0.00031775129325526573, "loss": 1.4912, "step": 7576 }, { "epoch": 0.75, "learning_rate": 0.0003175178813710586, "loss": 1.3691, "step": 7577 }, { "epoch": 0.75, "learning_rate": 0.0003172845390676001, "loss": 1.293, "step": 7578 }, { "epoch": 0.75, "learning_rate": 0.0003170512663686802, "loss": 1.332, "step": 7579 }, { "epoch": 0.75, "learning_rate": 0.0003168180632980816, "loss": 1.46, "step": 7580 }, { "epoch": 0.75, "learning_rate": 0.0003165849298795799, "loss": 1.2534, "step": 7581 }, { "epoch": 0.75, "learning_rate": 0.00031635186613694367, "loss": 1.3926, "step": 7582 }, { "epoch": 0.75, "learning_rate": 0.0003161188720939343, "loss": 1.3184, "step": 7583 }, { "epoch": 0.75, "learning_rate": 0.0003158859477743061, "loss": 1.4668, "step": 7584 }, { "epoch": 0.75, "learning_rate": 0.00031565309320180635, "loss": 1.3008, "step": 7585 }, { "epoch": 0.75, "learning_rate": 0.0003154203084001752, "loss": 1.3516, "step": 7586 }, { "epoch": 0.75, "learning_rate": 0.0003151875933931454, "loss": 1.4951, "step": 7587 }, { "epoch": 0.75, "learning_rate": 0.0003149549482044431, "loss": 1.373, "step": 7588 }, { "epoch": 0.75, "learning_rate": 0.00031472237285778694, "loss": 1.3779, "step": 7589 }, { "epoch": 0.75, "learning_rate": 0.00031448986737688846, "loss": 1.4258, "step": 7590 }, { "epoch": 0.75, "learning_rate": 0.0003142574317854523, "loss": 1.293, "step": 7591 }, { "epoch": 0.75, "learning_rate": 0.00031402506610717583, "loss": 1.4385, "step": 7592 }, { "epoch": 0.75, "learning_rate": 0.0003137927703657493, "loss": 1.335, "step": 7593 }, { "epoch": 0.75, "learning_rate": 0.0003135605445848557, "loss": 1.4443, "step": 7594 }, { "epoch": 0.75, "learning_rate": 0.0003133283887881713, "loss": 1.4443, "step": 7595 }, { "epoch": 0.75, "learning_rate": 0.00031309630299936476, "loss": 1.417, "step": 7596 }, { "epoch": 0.75, "learning_rate": 0.00031286428724209794, "loss": 1.3545, "step": 7597 }, { "epoch": 0.75, "learning_rate": 0.0003126323415400252, "loss": 1.3413, "step": 7598 }, { "epoch": 0.75, "learning_rate": 0.00031240046591679414, "loss": 1.4229, "step": 7599 }, { "epoch": 0.75, "learning_rate": 0.00031216866039604507, "loss": 1.4453, "step": 7600 }, { "epoch": 0.75, "learning_rate": 0.00031193692500141115, "loss": 1.3945, "step": 7601 }, { "epoch": 0.75, "learning_rate": 0.00031170525975651834, "loss": 1.4551, "step": 7602 }, { "epoch": 0.75, "learning_rate": 0.0003114736646849854, "loss": 1.4678, "step": 7603 }, { "epoch": 0.75, "learning_rate": 0.0003112421398104238, "loss": 1.3398, "step": 7604 }, { "epoch": 0.75, "learning_rate": 0.0003110106851564387, "loss": 1.2788, "step": 7605 }, { "epoch": 0.75, "learning_rate": 0.0003107793007466272, "loss": 1.2764, "step": 7606 }, { "epoch": 0.75, "learning_rate": 0.0003105479866045794, "loss": 1.396, "step": 7607 }, { "epoch": 0.75, "learning_rate": 0.00031031674275387857, "loss": 1.4102, "step": 7608 }, { "epoch": 0.75, "learning_rate": 0.00031008556921810026, "loss": 1.457, "step": 7609 }, { "epoch": 0.75, "learning_rate": 0.00030985446602081355, "loss": 1.4033, "step": 7610 }, { "epoch": 0.75, "learning_rate": 0.0003096234331855797, "loss": 1.4404, "step": 7611 }, { "epoch": 0.75, "learning_rate": 0.0003093924707359532, "loss": 1.3789, "step": 7612 }, { "epoch": 0.75, "learning_rate": 0.00030916157869548136, "loss": 1.4971, "step": 7613 }, { "epoch": 0.75, "learning_rate": 0.000308930757087704, "loss": 1.3584, "step": 7614 }, { "epoch": 0.75, "learning_rate": 0.0003087000059361541, "loss": 1.5034, "step": 7615 }, { "epoch": 0.75, "learning_rate": 0.00030846932526435713, "loss": 1.4971, "step": 7616 }, { "epoch": 0.75, "learning_rate": 0.0003082387150958318, "loss": 1.4941, "step": 7617 }, { "epoch": 0.75, "learning_rate": 0.0003080081754540892, "loss": 1.3057, "step": 7618 }, { "epoch": 0.75, "learning_rate": 0.0003077777063626335, "loss": 1.4844, "step": 7619 }, { "epoch": 0.75, "learning_rate": 0.0003075473078449615, "loss": 1.5391, "step": 7620 }, { "epoch": 0.75, "learning_rate": 0.00030731697992456307, "loss": 1.5361, "step": 7621 }, { "epoch": 0.75, "learning_rate": 0.00030708672262492044, "loss": 1.4053, "step": 7622 }, { "epoch": 0.75, "learning_rate": 0.0003068565359695092, "loss": 1.5195, "step": 7623 }, { "epoch": 0.75, "learning_rate": 0.0003066264199817975, "loss": 1.4795, "step": 7624 }, { "epoch": 0.75, "learning_rate": 0.00030639637468524606, "loss": 1.4775, "step": 7625 }, { "epoch": 0.75, "learning_rate": 0.0003061664001033085, "loss": 1.3418, "step": 7626 }, { "epoch": 0.75, "learning_rate": 0.00030593649625943153, "loss": 1.4023, "step": 7627 }, { "epoch": 0.75, "learning_rate": 0.00030570666317705433, "loss": 1.4268, "step": 7628 }, { "epoch": 0.75, "learning_rate": 0.00030547690087960876, "loss": 1.4551, "step": 7629 }, { "epoch": 0.75, "learning_rate": 0.00030524720939052, "loss": 1.4795, "step": 7630 }, { "epoch": 0.75, "learning_rate": 0.0003050175887332054, "loss": 1.4141, "step": 7631 }, { "epoch": 0.75, "learning_rate": 0.00030478803893107544, "loss": 1.3467, "step": 7632 }, { "epoch": 0.75, "learning_rate": 0.00030455856000753326, "loss": 1.4961, "step": 7633 }, { "epoch": 0.75, "learning_rate": 0.0003043291519859748, "loss": 1.3623, "step": 7634 }, { "epoch": 0.75, "learning_rate": 0.0003040998148897889, "loss": 1.4302, "step": 7635 }, { "epoch": 0.75, "learning_rate": 0.00030387054874235687, "loss": 1.5488, "step": 7636 }, { "epoch": 0.75, "learning_rate": 0.00030364135356705313, "loss": 1.54, "step": 7637 }, { "epoch": 0.75, "learning_rate": 0.0003034122293872447, "loss": 1.4736, "step": 7638 }, { "epoch": 0.75, "learning_rate": 0.0003031831762262912, "loss": 1.5225, "step": 7639 }, { "epoch": 0.75, "learning_rate": 0.0003029541941075449, "loss": 1.4033, "step": 7640 }, { "epoch": 0.75, "learning_rate": 0.0003027252830543519, "loss": 1.4229, "step": 7641 }, { "epoch": 0.75, "learning_rate": 0.00030249644309004965, "loss": 1.377, "step": 7642 }, { "epoch": 0.75, "learning_rate": 0.00030226767423796954, "loss": 1.2886, "step": 7643 }, { "epoch": 0.75, "learning_rate": 0.0003020389765214343, "loss": 1.3169, "step": 7644 }, { "epoch": 0.75, "learning_rate": 0.00030181034996376057, "loss": 1.3379, "step": 7645 }, { "epoch": 0.75, "learning_rate": 0.00030158179458825753, "loss": 1.4688, "step": 7646 }, { "epoch": 0.75, "learning_rate": 0.0003013533104182269, "loss": 1.4619, "step": 7647 }, { "epoch": 0.75, "learning_rate": 0.00030112489747696313, "loss": 1.3906, "step": 7648 }, { "epoch": 0.75, "learning_rate": 0.00030089655578775364, "loss": 1.4072, "step": 7649 }, { "epoch": 0.75, "learning_rate": 0.00030066828537387826, "loss": 1.3281, "step": 7650 }, { "epoch": 0.75, "learning_rate": 0.00030044008625860986, "loss": 1.3818, "step": 7651 }, { "epoch": 0.75, "learning_rate": 0.00030021195846521376, "loss": 1.2998, "step": 7652 }, { "epoch": 0.75, "learning_rate": 0.00029998390201694836, "loss": 1.2441, "step": 7653 }, { "epoch": 0.75, "learning_rate": 0.00029975591693706417, "loss": 1.4941, "step": 7654 }, { "epoch": 0.75, "learning_rate": 0.0002995280032488054, "loss": 1.3281, "step": 7655 }, { "epoch": 0.75, "learning_rate": 0.00029930016097540814, "loss": 1.2822, "step": 7656 }, { "epoch": 0.75, "learning_rate": 0.0002990723901401016, "loss": 1.3711, "step": 7657 }, { "epoch": 0.75, "learning_rate": 0.00029884469076610733, "loss": 1.3877, "step": 7658 }, { "epoch": 0.76, "learning_rate": 0.00029861706287664003, "loss": 1.4902, "step": 7659 }, { "epoch": 0.76, "learning_rate": 0.0002983895064949069, "loss": 1.4912, "step": 7660 }, { "epoch": 0.76, "learning_rate": 0.000298162021644108, "loss": 1.46, "step": 7661 }, { "epoch": 0.76, "learning_rate": 0.0002979346083474357, "loss": 1.4121, "step": 7662 }, { "epoch": 0.76, "learning_rate": 0.00029770726662807555, "loss": 1.459, "step": 7663 }, { "epoch": 0.76, "learning_rate": 0.0002974799965092055, "loss": 1.4033, "step": 7664 }, { "epoch": 0.76, "learning_rate": 0.00029725279801399643, "loss": 1.355, "step": 7665 }, { "epoch": 0.76, "learning_rate": 0.0002970256711656117, "loss": 1.3809, "step": 7666 }, { "epoch": 0.76, "learning_rate": 0.00029679861598720745, "loss": 1.4629, "step": 7667 }, { "epoch": 0.76, "learning_rate": 0.00029657163250193253, "loss": 1.332, "step": 7668 }, { "epoch": 0.76, "learning_rate": 0.0002963447207329285, "loss": 1.4824, "step": 7669 }, { "epoch": 0.76, "learning_rate": 0.0002961178807033296, "loss": 1.5205, "step": 7670 }, { "epoch": 0.76, "learning_rate": 0.00029589111243626267, "loss": 1.4668, "step": 7671 }, { "epoch": 0.76, "learning_rate": 0.00029566441595484707, "loss": 1.3262, "step": 7672 }, { "epoch": 0.76, "learning_rate": 0.0002954377912821956, "loss": 1.4121, "step": 7673 }, { "epoch": 0.76, "learning_rate": 0.00029521123844141286, "loss": 1.333, "step": 7674 }, { "epoch": 0.76, "learning_rate": 0.0002949847574555966, "loss": 1.3857, "step": 7675 }, { "epoch": 0.76, "learning_rate": 0.0002947583483478371, "loss": 1.4033, "step": 7676 }, { "epoch": 0.76, "learning_rate": 0.0002945320111412173, "loss": 1.46, "step": 7677 }, { "epoch": 0.76, "learning_rate": 0.0002943057458588129, "loss": 1.459, "step": 7678 }, { "epoch": 0.76, "learning_rate": 0.000294079552523692, "loss": 1.499, "step": 7679 }, { "epoch": 0.76, "learning_rate": 0.0002938534311589158, "loss": 1.4766, "step": 7680 }, { "epoch": 0.76, "learning_rate": 0.0002936273817875379, "loss": 1.4365, "step": 7681 }, { "epoch": 0.76, "learning_rate": 0.00029340140443260453, "loss": 1.376, "step": 7682 }, { "epoch": 0.76, "learning_rate": 0.0002931754991171547, "loss": 1.3994, "step": 7683 }, { "epoch": 0.76, "learning_rate": 0.0002929496658642199, "loss": 1.457, "step": 7684 }, { "epoch": 0.76, "learning_rate": 0.0002927239046968245, "loss": 1.4307, "step": 7685 }, { "epoch": 0.76, "learning_rate": 0.0002924982156379855, "loss": 1.4248, "step": 7686 }, { "epoch": 0.76, "learning_rate": 0.0002922725987107122, "loss": 1.333, "step": 7687 }, { "epoch": 0.76, "learning_rate": 0.00029204705393800714, "loss": 1.5088, "step": 7688 }, { "epoch": 0.76, "learning_rate": 0.00029182158134286485, "loss": 1.311, "step": 7689 }, { "epoch": 0.76, "learning_rate": 0.0002915961809482727, "loss": 1.3301, "step": 7690 }, { "epoch": 0.76, "learning_rate": 0.0002913708527772113, "loss": 1.5391, "step": 7691 }, { "epoch": 0.76, "learning_rate": 0.00029114559685265327, "loss": 1.5098, "step": 7692 }, { "epoch": 0.76, "learning_rate": 0.0002909204131975639, "loss": 1.3984, "step": 7693 }, { "epoch": 0.76, "learning_rate": 0.00029069530183490136, "loss": 1.4482, "step": 7694 }, { "epoch": 0.76, "learning_rate": 0.00029047026278761613, "loss": 1.3223, "step": 7695 }, { "epoch": 0.76, "learning_rate": 0.00029024529607865157, "loss": 1.416, "step": 7696 }, { "epoch": 0.76, "learning_rate": 0.0002900204017309437, "loss": 1.5381, "step": 7697 }, { "epoch": 0.76, "learning_rate": 0.0002897955797674209, "loss": 1.4634, "step": 7698 }, { "epoch": 0.76, "learning_rate": 0.0002895708302110047, "loss": 1.3457, "step": 7699 }, { "epoch": 0.76, "learning_rate": 0.0002893461530846083, "loss": 1.3809, "step": 7700 }, { "epoch": 0.76, "learning_rate": 0.00028912154841113837, "loss": 1.4912, "step": 7701 }, { "epoch": 0.76, "learning_rate": 0.00028889701621349385, "loss": 1.377, "step": 7702 }, { "epoch": 0.76, "learning_rate": 0.0002886725565145665, "loss": 1.3848, "step": 7703 }, { "epoch": 0.76, "learning_rate": 0.00028844816933724004, "loss": 1.3975, "step": 7704 }, { "epoch": 0.76, "learning_rate": 0.00028822385470439216, "loss": 1.3613, "step": 7705 }, { "epoch": 0.76, "learning_rate": 0.00028799961263889176, "loss": 1.6182, "step": 7706 }, { "epoch": 0.76, "learning_rate": 0.0002877754431636011, "loss": 1.4199, "step": 7707 }, { "epoch": 0.76, "learning_rate": 0.0002875513463013745, "loss": 1.5156, "step": 7708 }, { "epoch": 0.76, "learning_rate": 0.0002873273220750594, "loss": 1.4336, "step": 7709 }, { "epoch": 0.76, "learning_rate": 0.0002871033705074957, "loss": 1.2812, "step": 7710 }, { "epoch": 0.76, "learning_rate": 0.0002868794916215157, "loss": 1.2676, "step": 7711 }, { "epoch": 0.76, "learning_rate": 0.00028665568543994447, "loss": 1.2803, "step": 7712 }, { "epoch": 0.76, "learning_rate": 0.0002864319519855995, "loss": 1.4785, "step": 7713 }, { "epoch": 0.76, "learning_rate": 0.0002862082912812911, "loss": 1.3828, "step": 7714 }, { "epoch": 0.76, "learning_rate": 0.00028598470334982194, "loss": 1.4434, "step": 7715 }, { "epoch": 0.76, "learning_rate": 0.00028576118821398746, "loss": 1.3506, "step": 7716 }, { "epoch": 0.76, "learning_rate": 0.00028553774589657545, "loss": 1.6406, "step": 7717 }, { "epoch": 0.76, "learning_rate": 0.0002853143764203665, "loss": 1.3555, "step": 7718 }, { "epoch": 0.76, "learning_rate": 0.00028509107980813364, "loss": 1.4033, "step": 7719 }, { "epoch": 0.76, "learning_rate": 0.00028486785608264266, "loss": 1.3071, "step": 7720 }, { "epoch": 0.76, "learning_rate": 0.0002846447052666515, "loss": 1.3496, "step": 7721 }, { "epoch": 0.76, "learning_rate": 0.0002844216273829109, "loss": 1.4937, "step": 7722 }, { "epoch": 0.76, "learning_rate": 0.0002841986224541647, "loss": 1.5146, "step": 7723 }, { "epoch": 0.76, "learning_rate": 0.00028397569050314853, "loss": 1.3682, "step": 7724 }, { "epoch": 0.76, "learning_rate": 0.0002837528315525908, "loss": 1.3857, "step": 7725 }, { "epoch": 0.76, "learning_rate": 0.0002835300456252126, "loss": 1.2891, "step": 7726 }, { "epoch": 0.76, "learning_rate": 0.00028330733274372743, "loss": 1.3936, "step": 7727 }, { "epoch": 0.76, "learning_rate": 0.00028308469293084147, "loss": 1.4922, "step": 7728 }, { "epoch": 0.76, "learning_rate": 0.00028286212620925343, "loss": 1.3652, "step": 7729 }, { "epoch": 0.76, "learning_rate": 0.0002826396326016545, "loss": 1.4863, "step": 7730 }, { "epoch": 0.76, "learning_rate": 0.0002824172121307285, "loss": 1.4365, "step": 7731 }, { "epoch": 0.76, "learning_rate": 0.00028219486481915166, "loss": 1.3438, "step": 7732 }, { "epoch": 0.76, "learning_rate": 0.000281972590689593, "loss": 1.4395, "step": 7733 }, { "epoch": 0.76, "learning_rate": 0.00028175038976471377, "loss": 1.4414, "step": 7734 }, { "epoch": 0.76, "learning_rate": 0.00028152826206716795, "loss": 1.4277, "step": 7735 }, { "epoch": 0.76, "learning_rate": 0.000281306207619602, "loss": 1.4512, "step": 7736 }, { "epoch": 0.76, "learning_rate": 0.00028108422644465504, "loss": 1.4805, "step": 7737 }, { "epoch": 0.76, "learning_rate": 0.00028086231856495837, "loss": 1.292, "step": 7738 }, { "epoch": 0.76, "learning_rate": 0.00028064048400313637, "loss": 1.4824, "step": 7739 }, { "epoch": 0.76, "learning_rate": 0.0002804187227818049, "loss": 1.4443, "step": 7740 }, { "epoch": 0.76, "learning_rate": 0.00028019703492357406, "loss": 1.4492, "step": 7741 }, { "epoch": 0.76, "learning_rate": 0.0002799754204510451, "loss": 1.4199, "step": 7742 }, { "epoch": 0.76, "learning_rate": 0.00027975387938681195, "loss": 1.375, "step": 7743 }, { "epoch": 0.76, "learning_rate": 0.00027953241175346144, "loss": 1.3975, "step": 7744 }, { "epoch": 0.76, "learning_rate": 0.0002793110175735728, "loss": 1.4941, "step": 7745 }, { "epoch": 0.76, "learning_rate": 0.00027908969686971753, "loss": 1.457, "step": 7746 }, { "epoch": 0.76, "learning_rate": 0.00027886844966446, "loss": 1.2754, "step": 7747 }, { "epoch": 0.76, "learning_rate": 0.00027864727598035687, "loss": 1.4971, "step": 7748 }, { "epoch": 0.76, "learning_rate": 0.0002784261758399572, "loss": 1.415, "step": 7749 }, { "epoch": 0.76, "learning_rate": 0.00027820514926580296, "loss": 1.4814, "step": 7750 }, { "epoch": 0.76, "learning_rate": 0.00027798419628042816, "loss": 1.4316, "step": 7751 }, { "epoch": 0.76, "learning_rate": 0.0002777633169063596, "loss": 1.4438, "step": 7752 }, { "epoch": 0.76, "learning_rate": 0.0002775425111661164, "loss": 1.4951, "step": 7753 }, { "epoch": 0.76, "learning_rate": 0.00027732177908221034, "loss": 1.501, "step": 7754 }, { "epoch": 0.76, "learning_rate": 0.00027710112067714566, "loss": 1.3359, "step": 7755 }, { "epoch": 0.76, "learning_rate": 0.00027688053597341877, "loss": 1.3413, "step": 7756 }, { "epoch": 0.76, "learning_rate": 0.00027666002499351905, "loss": 1.3789, "step": 7757 }, { "epoch": 0.76, "learning_rate": 0.00027643958775992814, "loss": 1.3779, "step": 7758 }, { "epoch": 0.76, "learning_rate": 0.00027621922429512015, "loss": 1.3457, "step": 7759 }, { "epoch": 0.76, "learning_rate": 0.0002759989346215616, "loss": 1.3916, "step": 7760 }, { "epoch": 0.77, "learning_rate": 0.00027577871876171157, "loss": 1.5117, "step": 7761 }, { "epoch": 0.77, "learning_rate": 0.0002755585767380218, "loss": 1.4512, "step": 7762 }, { "epoch": 0.77, "learning_rate": 0.00027533850857293607, "loss": 1.3506, "step": 7763 }, { "epoch": 0.77, "learning_rate": 0.000275118514288891, "loss": 1.4893, "step": 7764 }, { "epoch": 0.77, "learning_rate": 0.0002748985939083155, "loss": 1.3125, "step": 7765 }, { "epoch": 0.77, "learning_rate": 0.0002746787474536311, "loss": 1.4463, "step": 7766 }, { "epoch": 0.77, "learning_rate": 0.00027445897494725167, "loss": 1.3672, "step": 7767 }, { "epoch": 0.77, "learning_rate": 0.0002742392764115834, "loss": 1.4619, "step": 7768 }, { "epoch": 0.77, "learning_rate": 0.00027401965186902536, "loss": 1.5039, "step": 7769 }, { "epoch": 0.77, "learning_rate": 0.00027380010134196866, "loss": 1.4092, "step": 7770 }, { "epoch": 0.77, "learning_rate": 0.000273580624852797, "loss": 1.3994, "step": 7771 }, { "epoch": 0.77, "learning_rate": 0.0002733612224238867, "loss": 1.4565, "step": 7772 }, { "epoch": 0.77, "learning_rate": 0.0002731418940776059, "loss": 1.4111, "step": 7773 }, { "epoch": 0.77, "learning_rate": 0.0002729226398363164, "loss": 1.4307, "step": 7774 }, { "epoch": 0.77, "learning_rate": 0.0002727034597223713, "loss": 1.6152, "step": 7775 }, { "epoch": 0.77, "learning_rate": 0.00027248435375811663, "loss": 1.3486, "step": 7776 }, { "epoch": 0.77, "learning_rate": 0.00027226532196589096, "loss": 1.4502, "step": 7777 }, { "epoch": 0.77, "learning_rate": 0.00027204636436802476, "loss": 1.3364, "step": 7778 }, { "epoch": 0.77, "learning_rate": 0.00027182748098684154, "loss": 1.4863, "step": 7779 }, { "epoch": 0.77, "learning_rate": 0.000271608671844657, "loss": 1.377, "step": 7780 }, { "epoch": 0.77, "learning_rate": 0.0002713899369637792, "loss": 1.4561, "step": 7781 }, { "epoch": 0.77, "learning_rate": 0.00027117127636650873, "loss": 1.4062, "step": 7782 }, { "epoch": 0.77, "learning_rate": 0.00027095269007513866, "loss": 1.5352, "step": 7783 }, { "epoch": 0.77, "learning_rate": 0.00027073417811195434, "loss": 1.3916, "step": 7784 }, { "epoch": 0.77, "learning_rate": 0.0002705157404992335, "loss": 1.4336, "step": 7785 }, { "epoch": 0.77, "learning_rate": 0.00027029737725924654, "loss": 1.6475, "step": 7786 }, { "epoch": 0.77, "learning_rate": 0.0002700790884142561, "loss": 1.4688, "step": 7787 }, { "epoch": 0.77, "learning_rate": 0.00026986087398651735, "loss": 1.4502, "step": 7788 }, { "epoch": 0.77, "learning_rate": 0.00026964273399827764, "loss": 1.3232, "step": 7789 }, { "epoch": 0.77, "learning_rate": 0.00026942466847177694, "loss": 1.3613, "step": 7790 }, { "epoch": 0.77, "learning_rate": 0.0002692066774292473, "loss": 1.4336, "step": 7791 }, { "epoch": 0.77, "learning_rate": 0.000268988760892914, "loss": 1.4795, "step": 7792 }, { "epoch": 0.77, "learning_rate": 0.00026877091888499407, "loss": 1.3682, "step": 7793 }, { "epoch": 0.77, "learning_rate": 0.00026855315142769676, "loss": 1.3789, "step": 7794 }, { "epoch": 0.77, "learning_rate": 0.00026833545854322403, "loss": 1.4199, "step": 7795 }, { "epoch": 0.77, "learning_rate": 0.0002681178402537705, "loss": 1.3955, "step": 7796 }, { "epoch": 0.77, "learning_rate": 0.0002679002965815226, "loss": 1.4629, "step": 7797 }, { "epoch": 0.77, "learning_rate": 0.0002676828275486596, "loss": 1.4941, "step": 7798 }, { "epoch": 0.77, "learning_rate": 0.00026746543317735297, "loss": 1.3711, "step": 7799 }, { "epoch": 0.77, "learning_rate": 0.0002672481134897665, "loss": 1.418, "step": 7800 }, { "epoch": 0.77, "learning_rate": 0.0002670308685080567, "loss": 1.5059, "step": 7801 }, { "epoch": 0.77, "learning_rate": 0.000266813698254372, "loss": 1.3867, "step": 7802 }, { "epoch": 0.77, "learning_rate": 0.00026659660275085364, "loss": 1.3945, "step": 7803 }, { "epoch": 0.77, "learning_rate": 0.00026637958201963487, "loss": 1.4111, "step": 7804 }, { "epoch": 0.77, "learning_rate": 0.00026616263608284164, "loss": 1.5146, "step": 7805 }, { "epoch": 0.77, "learning_rate": 0.00026594576496259204, "loss": 1.4688, "step": 7806 }, { "epoch": 0.77, "learning_rate": 0.0002657289686809966, "loss": 1.4385, "step": 7807 }, { "epoch": 0.77, "learning_rate": 0.00026551224726015845, "loss": 1.3496, "step": 7808 }, { "epoch": 0.77, "learning_rate": 0.0002652956007221723, "loss": 1.4463, "step": 7809 }, { "epoch": 0.77, "learning_rate": 0.0002650790290891265, "loss": 1.4521, "step": 7810 }, { "epoch": 0.77, "learning_rate": 0.0002648625323831012, "loss": 1.5088, "step": 7811 }, { "epoch": 0.77, "learning_rate": 0.00026464611062616795, "loss": 1.3936, "step": 7812 }, { "epoch": 0.77, "learning_rate": 0.000264429763840392, "loss": 1.4131, "step": 7813 }, { "epoch": 0.77, "learning_rate": 0.0002642134920478303, "loss": 1.3418, "step": 7814 }, { "epoch": 0.77, "learning_rate": 0.0002639972952705323, "loss": 1.4209, "step": 7815 }, { "epoch": 0.77, "learning_rate": 0.00026378117353053986, "loss": 1.4277, "step": 7816 }, { "epoch": 0.77, "learning_rate": 0.0002635651268498871, "loss": 1.4658, "step": 7817 }, { "epoch": 0.77, "learning_rate": 0.00026334915525060053, "loss": 1.4385, "step": 7818 }, { "epoch": 0.77, "learning_rate": 0.00026313325875469907, "loss": 1.4395, "step": 7819 }, { "epoch": 0.77, "learning_rate": 0.00026291743738419374, "loss": 1.4004, "step": 7820 }, { "epoch": 0.77, "learning_rate": 0.0002627016911610881, "loss": 1.3984, "step": 7821 }, { "epoch": 0.77, "learning_rate": 0.0002624860201073781, "loss": 1.4014, "step": 7822 }, { "epoch": 0.77, "learning_rate": 0.0002622704242450516, "loss": 1.4629, "step": 7823 }, { "epoch": 0.77, "learning_rate": 0.0002620549035960897, "loss": 1.501, "step": 7824 }, { "epoch": 0.77, "learning_rate": 0.00026183945818246515, "loss": 1.3936, "step": 7825 }, { "epoch": 0.77, "learning_rate": 0.0002616240880261429, "loss": 1.4062, "step": 7826 }, { "epoch": 0.77, "learning_rate": 0.0002614087931490805, "loss": 1.3682, "step": 7827 }, { "epoch": 0.77, "learning_rate": 0.0002611935735732279, "loss": 1.4336, "step": 7828 }, { "epoch": 0.77, "learning_rate": 0.0002609784293205273, "loss": 1.3486, "step": 7829 }, { "epoch": 0.77, "learning_rate": 0.0002607633604129132, "loss": 1.4082, "step": 7830 }, { "epoch": 0.77, "learning_rate": 0.0002605483668723121, "loss": 1.2998, "step": 7831 }, { "epoch": 0.77, "learning_rate": 0.00026033344872064356, "loss": 1.418, "step": 7832 }, { "epoch": 0.77, "learning_rate": 0.0002601186059798187, "loss": 1.5176, "step": 7833 }, { "epoch": 0.77, "learning_rate": 0.0002599038386717415, "loss": 1.5117, "step": 7834 }, { "epoch": 0.77, "learning_rate": 0.0002596891468183078, "loss": 1.3633, "step": 7835 }, { "epoch": 0.77, "learning_rate": 0.00025947453044140623, "loss": 1.4707, "step": 7836 }, { "epoch": 0.77, "learning_rate": 0.0002592599895629172, "loss": 1.3916, "step": 7837 }, { "epoch": 0.77, "learning_rate": 0.00025904552420471383, "loss": 1.4541, "step": 7838 }, { "epoch": 0.77, "learning_rate": 0.00025883113438866133, "loss": 1.416, "step": 7839 }, { "epoch": 0.77, "learning_rate": 0.00025861682013661745, "loss": 1.3867, "step": 7840 }, { "epoch": 0.77, "learning_rate": 0.0002584025814704315, "loss": 1.3809, "step": 7841 }, { "epoch": 0.77, "learning_rate": 0.0002581884184119464, "loss": 1.3906, "step": 7842 }, { "epoch": 0.77, "learning_rate": 0.0002579743309829964, "loss": 1.3877, "step": 7843 }, { "epoch": 0.77, "learning_rate": 0.000257760319205408, "loss": 1.4287, "step": 7844 }, { "epoch": 0.77, "learning_rate": 0.0002575463831010003, "loss": 1.4365, "step": 7845 }, { "epoch": 0.77, "learning_rate": 0.0002573325226915848, "loss": 1.4688, "step": 7846 }, { "epoch": 0.77, "learning_rate": 0.0002571187379989649, "loss": 1.2969, "step": 7847 }, { "epoch": 0.77, "learning_rate": 0.0002569050290449365, "loss": 1.5508, "step": 7848 }, { "epoch": 0.77, "learning_rate": 0.000256691395851288, "loss": 1.3828, "step": 7849 }, { "epoch": 0.77, "learning_rate": 0.00025647783843979943, "loss": 1.498, "step": 7850 }, { "epoch": 0.77, "learning_rate": 0.0002562643568322438, "loss": 1.3672, "step": 7851 }, { "epoch": 0.77, "learning_rate": 0.00025605095105038614, "loss": 1.4482, "step": 7852 }, { "epoch": 0.77, "learning_rate": 0.00025583762111598343, "loss": 1.3789, "step": 7853 }, { "epoch": 0.77, "learning_rate": 0.0002556243670507853, "loss": 1.3906, "step": 7854 }, { "epoch": 0.77, "learning_rate": 0.0002554111888765336, "loss": 1.3311, "step": 7855 }, { "epoch": 0.77, "learning_rate": 0.00025519808661496237, "loss": 1.3389, "step": 7856 }, { "epoch": 0.77, "learning_rate": 0.00025498506028779786, "loss": 1.4912, "step": 7857 }, { "epoch": 0.77, "learning_rate": 0.0002547721099167586, "loss": 1.3535, "step": 7858 }, { "epoch": 0.77, "learning_rate": 0.0002545592355235552, "loss": 1.4434, "step": 7859 }, { "epoch": 0.77, "learning_rate": 0.0002543464371298914, "loss": 1.2642, "step": 7860 }, { "epoch": 0.77, "learning_rate": 0.00025413371475746196, "loss": 1.4131, "step": 7861 }, { "epoch": 0.78, "learning_rate": 0.0002539210684279546, "loss": 1.3896, "step": 7862 }, { "epoch": 0.78, "learning_rate": 0.0002537084981630493, "loss": 1.4858, "step": 7863 }, { "epoch": 0.78, "learning_rate": 0.00025349600398441786, "loss": 1.4502, "step": 7864 }, { "epoch": 0.78, "learning_rate": 0.0002532835859137248, "loss": 1.46, "step": 7865 }, { "epoch": 0.78, "learning_rate": 0.0002530712439726265, "loss": 1.4639, "step": 7866 }, { "epoch": 0.78, "learning_rate": 0.00025285897818277226, "loss": 1.4756, "step": 7867 }, { "epoch": 0.78, "learning_rate": 0.00025264678856580226, "loss": 1.4277, "step": 7868 }, { "epoch": 0.78, "learning_rate": 0.0002524346751433502, "loss": 1.4717, "step": 7869 }, { "epoch": 0.78, "learning_rate": 0.0002522226379370416, "loss": 1.4521, "step": 7870 }, { "epoch": 0.78, "learning_rate": 0.000252010676968494, "loss": 1.418, "step": 7871 }, { "epoch": 0.78, "learning_rate": 0.0002517987922593176, "loss": 1.3271, "step": 7872 }, { "epoch": 0.78, "learning_rate": 0.0002515869838311141, "loss": 1.4287, "step": 7873 }, { "epoch": 0.78, "learning_rate": 0.0002513752517054786, "loss": 1.3613, "step": 7874 }, { "epoch": 0.78, "learning_rate": 0.0002511635959039973, "loss": 1.5205, "step": 7875 }, { "epoch": 0.78, "learning_rate": 0.00025095201644824915, "loss": 1.5684, "step": 7876 }, { "epoch": 0.78, "learning_rate": 0.0002507405133598051, "loss": 1.4639, "step": 7877 }, { "epoch": 0.78, "learning_rate": 0.00025052908666022846, "loss": 1.3584, "step": 7878 }, { "epoch": 0.78, "learning_rate": 0.0002503177363710747, "loss": 1.3447, "step": 7879 }, { "epoch": 0.78, "learning_rate": 0.0002501064625138916, "loss": 1.5088, "step": 7880 }, { "epoch": 0.78, "learning_rate": 0.00024989526511021886, "loss": 1.4902, "step": 7881 }, { "epoch": 0.78, "learning_rate": 0.0002496841441815888, "loss": 1.3525, "step": 7882 }, { "epoch": 0.78, "learning_rate": 0.0002494730997495256, "loss": 1.6025, "step": 7883 }, { "epoch": 0.78, "learning_rate": 0.00024926213183554567, "loss": 1.4404, "step": 7884 }, { "epoch": 0.78, "learning_rate": 0.0002490512404611579, "loss": 1.2598, "step": 7885 }, { "epoch": 0.78, "learning_rate": 0.00024884042564786316, "loss": 1.3994, "step": 7886 }, { "epoch": 0.78, "learning_rate": 0.0002486296874171544, "loss": 1.3525, "step": 7887 }, { "epoch": 0.78, "learning_rate": 0.0002484190257905171, "loss": 1.3975, "step": 7888 }, { "epoch": 0.78, "learning_rate": 0.0002482084407894286, "loss": 1.3477, "step": 7889 }, { "epoch": 0.78, "learning_rate": 0.0002479979324353586, "loss": 1.3066, "step": 7890 }, { "epoch": 0.78, "learning_rate": 0.00024778750074976876, "loss": 1.3984, "step": 7891 }, { "epoch": 0.78, "learning_rate": 0.0002475771457541135, "loss": 1.4082, "step": 7892 }, { "epoch": 0.78, "learning_rate": 0.00024736686746983886, "loss": 1.3779, "step": 7893 }, { "epoch": 0.78, "learning_rate": 0.00024715666591838316, "loss": 1.3779, "step": 7894 }, { "epoch": 0.78, "learning_rate": 0.0002469465411211771, "loss": 1.4443, "step": 7895 }, { "epoch": 0.78, "learning_rate": 0.0002467364930996434, "loss": 1.3589, "step": 7896 }, { "epoch": 0.78, "learning_rate": 0.00024652652187519687, "loss": 1.5723, "step": 7897 }, { "epoch": 0.78, "learning_rate": 0.00024631662746924464, "loss": 1.501, "step": 7898 }, { "epoch": 0.78, "learning_rate": 0.000246106809903186, "loss": 1.3818, "step": 7899 }, { "epoch": 0.78, "learning_rate": 0.0002458970691984125, "loss": 1.3252, "step": 7900 }, { "epoch": 0.78, "learning_rate": 0.0002456874053763074, "loss": 1.459, "step": 7901 }, { "epoch": 0.78, "learning_rate": 0.0002454778184582467, "loss": 1.4688, "step": 7902 }, { "epoch": 0.78, "learning_rate": 0.0002452683084655982, "loss": 1.4004, "step": 7903 }, { "epoch": 0.78, "learning_rate": 0.00024505887541972205, "loss": 1.4834, "step": 7904 }, { "epoch": 0.78, "learning_rate": 0.0002448495193419705, "loss": 1.4824, "step": 7905 }, { "epoch": 0.78, "learning_rate": 0.0002446402402536878, "loss": 1.4971, "step": 7906 }, { "epoch": 0.78, "learning_rate": 0.0002444310381762105, "loss": 1.4463, "step": 7907 }, { "epoch": 0.78, "learning_rate": 0.0002442219131308673, "loss": 1.4844, "step": 7908 }, { "epoch": 0.78, "learning_rate": 0.00024401286513897868, "loss": 1.3809, "step": 7909 }, { "epoch": 0.78, "learning_rate": 0.00024380389422185822, "loss": 1.5195, "step": 7910 }, { "epoch": 0.78, "learning_rate": 0.00024359500040081083, "loss": 1.3848, "step": 7911 }, { "epoch": 0.78, "learning_rate": 0.0002433861836971336, "loss": 1.3096, "step": 7912 }, { "epoch": 0.78, "learning_rate": 0.00024317744413211595, "loss": 1.4727, "step": 7913 }, { "epoch": 0.78, "learning_rate": 0.0002429687817270395, "loss": 1.4932, "step": 7914 }, { "epoch": 0.78, "learning_rate": 0.0002427601965031778, "loss": 1.3271, "step": 7915 }, { "epoch": 0.78, "learning_rate": 0.00024255168848179664, "loss": 1.5039, "step": 7916 }, { "epoch": 0.78, "learning_rate": 0.000242343257684154, "loss": 1.4062, "step": 7917 }, { "epoch": 0.78, "learning_rate": 0.00024213490413149975, "loss": 1.415, "step": 7918 }, { "epoch": 0.78, "learning_rate": 0.00024192662784507625, "loss": 1.292, "step": 7919 }, { "epoch": 0.78, "learning_rate": 0.00024171842884611772, "loss": 1.4727, "step": 7920 }, { "epoch": 0.78, "learning_rate": 0.00024151030715585054, "loss": 1.3438, "step": 7921 }, { "epoch": 0.78, "learning_rate": 0.00024130226279549317, "loss": 1.3955, "step": 7922 }, { "epoch": 0.78, "learning_rate": 0.0002410942957862564, "loss": 1.5713, "step": 7923 }, { "epoch": 0.78, "learning_rate": 0.00024088640614934286, "loss": 1.3843, "step": 7924 }, { "epoch": 0.78, "learning_rate": 0.00024067859390594748, "loss": 1.4443, "step": 7925 }, { "epoch": 0.78, "learning_rate": 0.00024047085907725717, "loss": 1.2607, "step": 7926 }, { "epoch": 0.78, "learning_rate": 0.00024026320168445115, "loss": 1.4355, "step": 7927 }, { "epoch": 0.78, "learning_rate": 0.00024005562174870054, "loss": 1.3975, "step": 7928 }, { "epoch": 0.78, "learning_rate": 0.00023984811929116847, "loss": 1.4473, "step": 7929 }, { "epoch": 0.78, "learning_rate": 0.0002396406943330106, "loss": 1.4272, "step": 7930 }, { "epoch": 0.78, "learning_rate": 0.0002394333468953742, "loss": 1.417, "step": 7931 }, { "epoch": 0.78, "learning_rate": 0.00023922607699939903, "loss": 1.377, "step": 7932 }, { "epoch": 0.78, "learning_rate": 0.00023901888466621658, "loss": 1.4258, "step": 7933 }, { "epoch": 0.78, "learning_rate": 0.00023881176991695074, "loss": 1.4521, "step": 7934 }, { "epoch": 0.78, "learning_rate": 0.00023860473277271743, "loss": 1.2979, "step": 7935 }, { "epoch": 0.78, "learning_rate": 0.00023839777325462453, "loss": 1.3926, "step": 7936 }, { "epoch": 0.78, "learning_rate": 0.00023819089138377205, "loss": 1.4473, "step": 7937 }, { "epoch": 0.78, "learning_rate": 0.00023798408718125218, "loss": 1.3818, "step": 7938 }, { "epoch": 0.78, "learning_rate": 0.00023777736066814904, "loss": 1.3379, "step": 7939 }, { "epoch": 0.78, "learning_rate": 0.00023757071186553903, "loss": 1.3828, "step": 7940 }, { "epoch": 0.78, "learning_rate": 0.00023736414079449008, "loss": 1.3271, "step": 7941 }, { "epoch": 0.78, "learning_rate": 0.00023715764747606326, "loss": 1.3975, "step": 7942 }, { "epoch": 0.78, "learning_rate": 0.00023695123193131086, "loss": 1.4316, "step": 7943 }, { "epoch": 0.78, "learning_rate": 0.0002367448941812773, "loss": 1.5029, "step": 7944 }, { "epoch": 0.78, "learning_rate": 0.0002365386342469994, "loss": 1.5352, "step": 7945 }, { "epoch": 0.78, "learning_rate": 0.00023633245214950584, "loss": 1.3779, "step": 7946 }, { "epoch": 0.78, "learning_rate": 0.00023612634790981724, "loss": 1.4189, "step": 7947 }, { "epoch": 0.78, "learning_rate": 0.0002359203215489466, "loss": 1.3755, "step": 7948 }, { "epoch": 0.78, "learning_rate": 0.00023571437308789878, "loss": 1.4141, "step": 7949 }, { "epoch": 0.78, "learning_rate": 0.00023550850254767076, "loss": 1.292, "step": 7950 }, { "epoch": 0.78, "learning_rate": 0.00023530270994925152, "loss": 1.5215, "step": 7951 }, { "epoch": 0.78, "learning_rate": 0.00023509699531362206, "loss": 1.3916, "step": 7952 }, { "epoch": 0.78, "learning_rate": 0.0002348913586617556, "loss": 1.4443, "step": 7953 }, { "epoch": 0.78, "learning_rate": 0.00023468580001461737, "loss": 1.4307, "step": 7954 }, { "epoch": 0.78, "learning_rate": 0.0002344803193931644, "loss": 1.4199, "step": 7955 }, { "epoch": 0.78, "learning_rate": 0.00023427491681834601, "loss": 1.4214, "step": 7956 }, { "epoch": 0.78, "learning_rate": 0.0002340695923111036, "loss": 1.4502, "step": 7957 }, { "epoch": 0.78, "learning_rate": 0.0002338643458923705, "loss": 1.3779, "step": 7958 }, { "epoch": 0.78, "learning_rate": 0.00023365917758307164, "loss": 1.4668, "step": 7959 }, { "epoch": 0.78, "learning_rate": 0.00023345408740412522, "loss": 1.416, "step": 7960 }, { "epoch": 0.78, "learning_rate": 0.00023324907537644024, "loss": 1.4316, "step": 7961 }, { "epoch": 0.78, "learning_rate": 0.00023304414152091834, "loss": 1.4482, "step": 7962 }, { "epoch": 0.78, "learning_rate": 0.00023283928585845294, "loss": 1.3838, "step": 7963 }, { "epoch": 0.79, "learning_rate": 0.0002326345084099295, "loss": 1.4551, "step": 7964 }, { "epoch": 0.79, "learning_rate": 0.0002324298091962258, "loss": 1.3682, "step": 7965 }, { "epoch": 0.79, "learning_rate": 0.00023222518823821125, "loss": 1.4844, "step": 7966 }, { "epoch": 0.79, "learning_rate": 0.00023202064555674762, "loss": 1.2666, "step": 7967 }, { "epoch": 0.79, "learning_rate": 0.00023181618117268843, "loss": 1.502, "step": 7968 }, { "epoch": 0.79, "learning_rate": 0.00023161179510687935, "loss": 1.3945, "step": 7969 }, { "epoch": 0.79, "learning_rate": 0.00023140748738015817, "loss": 1.4053, "step": 7970 }, { "epoch": 0.79, "learning_rate": 0.00023120325801335428, "loss": 1.4238, "step": 7971 }, { "epoch": 0.79, "learning_rate": 0.00023099910702728966, "loss": 1.3652, "step": 7972 }, { "epoch": 0.79, "learning_rate": 0.00023079503444277795, "loss": 1.4512, "step": 7973 }, { "epoch": 0.79, "learning_rate": 0.00023059104028062472, "loss": 1.3271, "step": 7974 }, { "epoch": 0.79, "learning_rate": 0.00023038712456162791, "loss": 1.4414, "step": 7975 }, { "epoch": 0.79, "learning_rate": 0.000230183287306577, "loss": 1.5479, "step": 7976 }, { "epoch": 0.79, "learning_rate": 0.0002299795285362538, "loss": 1.5391, "step": 7977 }, { "epoch": 0.79, "learning_rate": 0.00022977584827143182, "loss": 1.3071, "step": 7978 }, { "epoch": 0.79, "learning_rate": 0.00022957224653287722, "loss": 1.3281, "step": 7979 }, { "epoch": 0.79, "learning_rate": 0.00022936872334134773, "loss": 1.4492, "step": 7980 }, { "epoch": 0.79, "learning_rate": 0.00022916527871759252, "loss": 1.4756, "step": 7981 }, { "epoch": 0.79, "learning_rate": 0.00022896191268235346, "loss": 1.4766, "step": 7982 }, { "epoch": 0.79, "learning_rate": 0.00022875862525636436, "loss": 1.4238, "step": 7983 }, { "epoch": 0.79, "learning_rate": 0.00022855541646035082, "loss": 1.4287, "step": 7984 }, { "epoch": 0.79, "learning_rate": 0.00022835228631503047, "loss": 1.4053, "step": 7985 }, { "epoch": 0.79, "learning_rate": 0.00022814923484111295, "loss": 1.4219, "step": 7986 }, { "epoch": 0.79, "learning_rate": 0.0002279462620592999, "loss": 1.3818, "step": 7987 }, { "epoch": 0.79, "learning_rate": 0.00022774336799028473, "loss": 1.4023, "step": 7988 }, { "epoch": 0.79, "learning_rate": 0.00022754055265475327, "loss": 1.4141, "step": 7989 }, { "epoch": 0.79, "learning_rate": 0.0002273378160733828, "loss": 1.459, "step": 7990 }, { "epoch": 0.79, "learning_rate": 0.00022713515826684283, "loss": 1.4961, "step": 7991 }, { "epoch": 0.79, "learning_rate": 0.0002269325792557948, "loss": 1.5645, "step": 7992 }, { "epoch": 0.79, "learning_rate": 0.00022673007906089237, "loss": 1.3374, "step": 7993 }, { "epoch": 0.79, "learning_rate": 0.00022652765770278093, "loss": 1.292, "step": 7994 }, { "epoch": 0.79, "learning_rate": 0.00022632531520209775, "loss": 1.3589, "step": 7995 }, { "epoch": 0.79, "learning_rate": 0.00022612305157947211, "loss": 1.374, "step": 7996 }, { "epoch": 0.79, "learning_rate": 0.00022592086685552527, "loss": 1.4863, "step": 7997 }, { "epoch": 0.79, "learning_rate": 0.0002257187610508705, "loss": 1.3135, "step": 7998 }, { "epoch": 0.79, "learning_rate": 0.00022551673418611307, "loss": 1.4854, "step": 7999 }, { "epoch": 0.79, "learning_rate": 0.00022531478628185, "loss": 1.3037, "step": 8000 }, { "epoch": 0.79, "learning_rate": 0.00022511291735867047, "loss": 1.4482, "step": 8001 }, { "epoch": 0.79, "learning_rate": 0.00022491112743715548, "loss": 1.4492, "step": 8002 }, { "epoch": 0.79, "learning_rate": 0.00022470941653787802, "loss": 1.4883, "step": 8003 }, { "epoch": 0.79, "learning_rate": 0.00022450778468140298, "loss": 1.4746, "step": 8004 }, { "epoch": 0.79, "learning_rate": 0.00022430623188828726, "loss": 1.418, "step": 8005 }, { "epoch": 0.79, "learning_rate": 0.0002241047581790796, "loss": 1.5068, "step": 8006 }, { "epoch": 0.79, "learning_rate": 0.00022390336357432095, "loss": 1.5537, "step": 8007 }, { "epoch": 0.79, "learning_rate": 0.00022370204809454375, "loss": 1.3545, "step": 8008 }, { "epoch": 0.79, "learning_rate": 0.0002235008117602727, "loss": 1.3232, "step": 8009 }, { "epoch": 0.79, "learning_rate": 0.00022329965459202416, "loss": 1.3574, "step": 8010 }, { "epoch": 0.79, "learning_rate": 0.00022309857661030708, "loss": 1.3438, "step": 8011 }, { "epoch": 0.79, "learning_rate": 0.00022289757783562159, "loss": 1.3555, "step": 8012 }, { "epoch": 0.79, "learning_rate": 0.00022269665828845997, "loss": 1.3174, "step": 8013 }, { "epoch": 0.79, "learning_rate": 0.00022249581798930651, "loss": 1.3379, "step": 8014 }, { "epoch": 0.79, "learning_rate": 0.0002222950569586374, "loss": 1.4141, "step": 8015 }, { "epoch": 0.79, "learning_rate": 0.00022209437521692068, "loss": 1.4805, "step": 8016 }, { "epoch": 0.79, "learning_rate": 0.00022189377278461643, "loss": 1.4736, "step": 8017 }, { "epoch": 0.79, "learning_rate": 0.0002216932496821764, "loss": 1.3945, "step": 8018 }, { "epoch": 0.79, "learning_rate": 0.00022149280593004462, "loss": 1.4219, "step": 8019 }, { "epoch": 0.79, "learning_rate": 0.0002212924415486568, "loss": 1.4932, "step": 8020 }, { "epoch": 0.79, "learning_rate": 0.00022109215655844038, "loss": 1.4795, "step": 8021 }, { "epoch": 0.79, "learning_rate": 0.00022089195097981518, "loss": 1.4395, "step": 8022 }, { "epoch": 0.79, "learning_rate": 0.00022069182483319238, "loss": 1.5264, "step": 8023 }, { "epoch": 0.79, "learning_rate": 0.00022049177813897569, "loss": 1.4414, "step": 8024 }, { "epoch": 0.79, "learning_rate": 0.00022029181091756, "loss": 1.4189, "step": 8025 }, { "epoch": 0.79, "learning_rate": 0.00022009192318933268, "loss": 1.3398, "step": 8026 }, { "epoch": 0.79, "learning_rate": 0.00021989211497467286, "loss": 1.3682, "step": 8027 }, { "epoch": 0.79, "learning_rate": 0.00021969238629395094, "loss": 1.3623, "step": 8028 }, { "epoch": 0.79, "learning_rate": 0.0002194927371675305, "loss": 1.3271, "step": 8029 }, { "epoch": 0.79, "learning_rate": 0.000219293167615766, "loss": 1.623, "step": 8030 }, { "epoch": 0.79, "learning_rate": 0.0002190936776590039, "loss": 1.2065, "step": 8031 }, { "epoch": 0.79, "learning_rate": 0.00021889426731758278, "loss": 1.5215, "step": 8032 }, { "epoch": 0.79, "learning_rate": 0.00021869493661183314, "loss": 1.418, "step": 8033 }, { "epoch": 0.79, "learning_rate": 0.0002184956855620771, "loss": 1.332, "step": 8034 }, { "epoch": 0.79, "learning_rate": 0.00021829651418862873, "loss": 1.397, "step": 8035 }, { "epoch": 0.79, "learning_rate": 0.00021809742251179454, "loss": 1.4785, "step": 8036 }, { "epoch": 0.79, "learning_rate": 0.00021789841055187178, "loss": 1.4473, "step": 8037 }, { "epoch": 0.79, "learning_rate": 0.00021769947832915026, "loss": 1.4287, "step": 8038 }, { "epoch": 0.79, "learning_rate": 0.00021750062586391205, "loss": 1.4473, "step": 8039 }, { "epoch": 0.79, "learning_rate": 0.00021730185317643026, "loss": 1.4756, "step": 8040 }, { "epoch": 0.79, "learning_rate": 0.00021710316028697053, "loss": 1.3535, "step": 8041 }, { "epoch": 0.79, "learning_rate": 0.00021690454721578966, "loss": 1.3701, "step": 8042 }, { "epoch": 0.79, "learning_rate": 0.0002167060139831374, "loss": 1.3848, "step": 8043 }, { "epoch": 0.79, "learning_rate": 0.00021650756060925426, "loss": 1.4014, "step": 8044 }, { "epoch": 0.79, "learning_rate": 0.00021630918711437332, "loss": 1.4219, "step": 8045 }, { "epoch": 0.79, "learning_rate": 0.00021611089351871894, "loss": 1.4277, "step": 8046 }, { "epoch": 0.79, "learning_rate": 0.00021591267984250796, "loss": 1.4346, "step": 8047 }, { "epoch": 0.79, "learning_rate": 0.00021571454610594854, "loss": 1.3145, "step": 8048 }, { "epoch": 0.79, "learning_rate": 0.00021551649232924098, "loss": 1.4141, "step": 8049 }, { "epoch": 0.79, "learning_rate": 0.00021531851853257732, "loss": 1.3584, "step": 8050 }, { "epoch": 0.79, "learning_rate": 0.00021512062473614158, "loss": 1.3555, "step": 8051 }, { "epoch": 0.79, "learning_rate": 0.00021492281096010935, "loss": 1.333, "step": 8052 }, { "epoch": 0.79, "learning_rate": 0.0002147250772246484, "loss": 1.377, "step": 8053 }, { "epoch": 0.79, "learning_rate": 0.00021452742354991818, "loss": 1.3584, "step": 8054 }, { "epoch": 0.79, "learning_rate": 0.00021432984995606986, "loss": 1.4326, "step": 8055 }, { "epoch": 0.79, "learning_rate": 0.00021413235646324658, "loss": 1.3955, "step": 8056 }, { "epoch": 0.79, "learning_rate": 0.00021393494309158336, "loss": 1.4429, "step": 8057 }, { "epoch": 0.79, "learning_rate": 0.00021373760986120684, "loss": 1.4009, "step": 8058 }, { "epoch": 0.79, "learning_rate": 0.00021354035679223583, "loss": 1.4463, "step": 8059 }, { "epoch": 0.79, "learning_rate": 0.00021334318390478046, "loss": 1.3838, "step": 8060 }, { "epoch": 0.79, "learning_rate": 0.00021314609121894334, "loss": 1.29, "step": 8061 }, { "epoch": 0.79, "learning_rate": 0.00021294907875481862, "loss": 1.4365, "step": 8062 }, { "epoch": 0.79, "learning_rate": 0.00021275214653249187, "loss": 1.5117, "step": 8063 }, { "epoch": 0.79, "learning_rate": 0.00021255529457204104, "loss": 1.3896, "step": 8064 }, { "epoch": 0.8, "learning_rate": 0.00021235852289353563, "loss": 1.4526, "step": 8065 }, { "epoch": 0.8, "learning_rate": 0.00021216183151703694, "loss": 1.4092, "step": 8066 }, { "epoch": 0.8, "learning_rate": 0.00021196522046259815, "loss": 1.5342, "step": 8067 }, { "epoch": 0.8, "learning_rate": 0.00021176868975026432, "loss": 1.3174, "step": 8068 }, { "epoch": 0.8, "learning_rate": 0.00021157223940007232, "loss": 1.2988, "step": 8069 }, { "epoch": 0.8, "learning_rate": 0.00021137586943205044, "loss": 1.3438, "step": 8070 }, { "epoch": 0.8, "learning_rate": 0.0002111795798662195, "loss": 1.4688, "step": 8071 }, { "epoch": 0.8, "learning_rate": 0.00021098337072259133, "loss": 1.3887, "step": 8072 }, { "epoch": 0.8, "learning_rate": 0.00021078724202117015, "loss": 1.3105, "step": 8073 }, { "epoch": 0.8, "learning_rate": 0.0002105911937819518, "loss": 1.4824, "step": 8074 }, { "epoch": 0.8, "learning_rate": 0.00021039522602492368, "loss": 1.5049, "step": 8075 }, { "epoch": 0.8, "learning_rate": 0.00021019933877006546, "loss": 1.3066, "step": 8076 }, { "epoch": 0.8, "learning_rate": 0.00021000353203734813, "loss": 1.3643, "step": 8077 }, { "epoch": 0.8, "learning_rate": 0.00020980780584673443, "loss": 1.4336, "step": 8078 }, { "epoch": 0.8, "learning_rate": 0.00020961216021817976, "loss": 1.3965, "step": 8079 }, { "epoch": 0.8, "learning_rate": 0.00020941659517163024, "loss": 1.498, "step": 8080 }, { "epoch": 0.8, "learning_rate": 0.00020922111072702444, "loss": 1.3809, "step": 8081 }, { "epoch": 0.8, "learning_rate": 0.00020902570690429234, "loss": 1.3691, "step": 8082 }, { "epoch": 0.8, "learning_rate": 0.00020883038372335595, "loss": 1.4521, "step": 8083 }, { "epoch": 0.8, "learning_rate": 0.00020863514120412886, "loss": 1.4209, "step": 8084 }, { "epoch": 0.8, "learning_rate": 0.00020843997936651647, "loss": 1.374, "step": 8085 }, { "epoch": 0.8, "learning_rate": 0.00020824489823041613, "loss": 1.4033, "step": 8086 }, { "epoch": 0.8, "learning_rate": 0.00020804989781571682, "loss": 1.3594, "step": 8087 }, { "epoch": 0.8, "learning_rate": 0.00020785497814229935, "loss": 1.3896, "step": 8088 }, { "epoch": 0.8, "learning_rate": 0.00020766013923003625, "loss": 1.3701, "step": 8089 }, { "epoch": 0.8, "learning_rate": 0.00020746538109879176, "loss": 1.3955, "step": 8090 }, { "epoch": 0.8, "learning_rate": 0.00020727070376842194, "loss": 1.417, "step": 8091 }, { "epoch": 0.8, "learning_rate": 0.0002070761072587748, "loss": 1.5078, "step": 8092 }, { "epoch": 0.8, "learning_rate": 0.00020688159158968977, "loss": 1.418, "step": 8093 }, { "epoch": 0.8, "learning_rate": 0.00020668715678099825, "loss": 1.46, "step": 8094 }, { "epoch": 0.8, "learning_rate": 0.00020649280285252346, "loss": 1.3296, "step": 8095 }, { "epoch": 0.8, "learning_rate": 0.00020629852982408014, "loss": 1.3242, "step": 8096 }, { "epoch": 0.8, "learning_rate": 0.00020610433771547487, "loss": 1.4199, "step": 8097 }, { "epoch": 0.8, "learning_rate": 0.00020591022654650615, "loss": 1.3545, "step": 8098 }, { "epoch": 0.8, "learning_rate": 0.00020571619633696404, "loss": 1.4902, "step": 8099 }, { "epoch": 0.8, "learning_rate": 0.00020552224710663038, "loss": 1.377, "step": 8100 }, { "epoch": 0.8, "learning_rate": 0.0002053283788752788, "loss": 1.3418, "step": 8101 }, { "epoch": 0.8, "learning_rate": 0.00020513459166267468, "loss": 1.2534, "step": 8102 }, { "epoch": 0.8, "learning_rate": 0.00020494088548857504, "loss": 1.373, "step": 8103 }, { "epoch": 0.8, "learning_rate": 0.00020474726037272872, "loss": 1.3262, "step": 8104 }, { "epoch": 0.8, "learning_rate": 0.0002045537163348763, "loss": 1.3169, "step": 8105 }, { "epoch": 0.8, "learning_rate": 0.00020436025339475006, "loss": 1.4004, "step": 8106 }, { "epoch": 0.8, "learning_rate": 0.00020416687157207415, "loss": 1.4102, "step": 8107 }, { "epoch": 0.8, "learning_rate": 0.0002039735708865641, "loss": 1.2925, "step": 8108 }, { "epoch": 0.8, "learning_rate": 0.00020378035135792762, "loss": 1.3057, "step": 8109 }, { "epoch": 0.8, "learning_rate": 0.00020358721300586346, "loss": 1.3818, "step": 8110 }, { "epoch": 0.8, "learning_rate": 0.00020339415585006327, "loss": 1.3584, "step": 8111 }, { "epoch": 0.8, "learning_rate": 0.00020320117991020926, "loss": 1.4321, "step": 8112 }, { "epoch": 0.8, "learning_rate": 0.0002030082852059759, "loss": 1.4404, "step": 8113 }, { "epoch": 0.8, "learning_rate": 0.0002028154717570293, "loss": 1.3438, "step": 8114 }, { "epoch": 0.8, "learning_rate": 0.00020262273958302735, "loss": 1.374, "step": 8115 }, { "epoch": 0.8, "learning_rate": 0.00020243008870361934, "loss": 1.4092, "step": 8116 }, { "epoch": 0.8, "learning_rate": 0.0002022375191384467, "loss": 1.4775, "step": 8117 }, { "epoch": 0.8, "learning_rate": 0.00020204503090714233, "loss": 1.4404, "step": 8118 }, { "epoch": 0.8, "learning_rate": 0.0002018526240293308, "loss": 1.4922, "step": 8119 }, { "epoch": 0.8, "learning_rate": 0.00020166029852462865, "loss": 1.417, "step": 8120 }, { "epoch": 0.8, "learning_rate": 0.00020146805441264383, "loss": 1.2578, "step": 8121 }, { "epoch": 0.8, "learning_rate": 0.000201275891712976, "loss": 1.4023, "step": 8122 }, { "epoch": 0.8, "learning_rate": 0.00020108381044521684, "loss": 1.3242, "step": 8123 }, { "epoch": 0.8, "learning_rate": 0.00020089181062894935, "loss": 1.333, "step": 8124 }, { "epoch": 0.8, "learning_rate": 0.00020069989228374842, "loss": 1.4541, "step": 8125 }, { "epoch": 0.8, "learning_rate": 0.0002005080554291807, "loss": 1.2783, "step": 8126 }, { "epoch": 0.8, "learning_rate": 0.00020031630008480428, "loss": 1.4717, "step": 8127 }, { "epoch": 0.8, "learning_rate": 0.00020012462627016904, "loss": 1.2759, "step": 8128 }, { "epoch": 0.8, "learning_rate": 0.00019993303400481688, "loss": 1.4609, "step": 8129 }, { "epoch": 0.8, "learning_rate": 0.00019974152330828098, "loss": 1.3516, "step": 8130 }, { "epoch": 0.8, "learning_rate": 0.0001995500942000864, "loss": 1.3848, "step": 8131 }, { "epoch": 0.8, "learning_rate": 0.00019935874669974973, "loss": 1.4512, "step": 8132 }, { "epoch": 0.8, "learning_rate": 0.00019916748082677926, "loss": 1.3027, "step": 8133 }, { "epoch": 0.8, "learning_rate": 0.00019897629660067508, "loss": 1.3486, "step": 8134 }, { "epoch": 0.8, "learning_rate": 0.00019878519404092898, "loss": 1.4922, "step": 8135 }, { "epoch": 0.8, "learning_rate": 0.00019859417316702422, "loss": 1.585, "step": 8136 }, { "epoch": 0.8, "learning_rate": 0.00019840323399843585, "loss": 1.2412, "step": 8137 }, { "epoch": 0.8, "learning_rate": 0.00019821237655463065, "loss": 1.5029, "step": 8138 }, { "epoch": 0.8, "learning_rate": 0.00019802160085506703, "loss": 1.4912, "step": 8139 }, { "epoch": 0.8, "learning_rate": 0.00019783090691919504, "loss": 1.4766, "step": 8140 }, { "epoch": 0.8, "learning_rate": 0.0001976402947664564, "loss": 1.3579, "step": 8141 }, { "epoch": 0.8, "learning_rate": 0.0001974497644162845, "loss": 1.3027, "step": 8142 }, { "epoch": 0.8, "learning_rate": 0.0001972593158881042, "loss": 1.4053, "step": 8143 }, { "epoch": 0.8, "learning_rate": 0.00019706894920133245, "loss": 1.4102, "step": 8144 }, { "epoch": 0.8, "learning_rate": 0.0001968786643753776, "loss": 1.3086, "step": 8145 }, { "epoch": 0.8, "learning_rate": 0.00019668846142963914, "loss": 1.3726, "step": 8146 }, { "epoch": 0.8, "learning_rate": 0.00019649834038350955, "loss": 1.3799, "step": 8147 }, { "epoch": 0.8, "learning_rate": 0.00019630830125637178, "loss": 1.4131, "step": 8148 }, { "epoch": 0.8, "learning_rate": 0.000196118344067601, "loss": 1.3184, "step": 8149 }, { "epoch": 0.8, "learning_rate": 0.00019592846883656324, "loss": 1.3994, "step": 8150 }, { "epoch": 0.8, "learning_rate": 0.00019573867558261715, "loss": 1.3164, "step": 8151 }, { "epoch": 0.8, "learning_rate": 0.0001955489643251126, "loss": 1.2612, "step": 8152 }, { "epoch": 0.8, "learning_rate": 0.00019535933508339109, "loss": 1.5322, "step": 8153 }, { "epoch": 0.8, "learning_rate": 0.00019516978787678584, "loss": 1.3018, "step": 8154 }, { "epoch": 0.8, "learning_rate": 0.00019498032272462164, "loss": 1.3579, "step": 8155 }, { "epoch": 0.8, "learning_rate": 0.00019479093964621486, "loss": 1.3105, "step": 8156 }, { "epoch": 0.8, "learning_rate": 0.00019460163866087354, "loss": 1.3433, "step": 8157 }, { "epoch": 0.8, "learning_rate": 0.0001944124197878976, "loss": 1.4238, "step": 8158 }, { "epoch": 0.8, "learning_rate": 0.00019422328304657821, "loss": 1.4551, "step": 8159 }, { "epoch": 0.8, "learning_rate": 0.00019403422845619812, "loss": 1.313, "step": 8160 }, { "epoch": 0.8, "learning_rate": 0.00019384525603603243, "loss": 1.3179, "step": 8161 }, { "epoch": 0.8, "learning_rate": 0.00019365636580534706, "loss": 1.3252, "step": 8162 }, { "epoch": 0.8, "learning_rate": 0.00019346755778339996, "loss": 1.3096, "step": 8163 }, { "epoch": 0.8, "learning_rate": 0.0001932788319894404, "loss": 1.3394, "step": 8164 }, { "epoch": 0.8, "learning_rate": 0.00019309018844270953, "loss": 1.3682, "step": 8165 }, { "epoch": 0.81, "learning_rate": 0.00019290162716244008, "loss": 1.4053, "step": 8166 }, { "epoch": 0.81, "learning_rate": 0.00019271314816785622, "loss": 1.4082, "step": 8167 }, { "epoch": 0.81, "learning_rate": 0.00019252475147817395, "loss": 1.3984, "step": 8168 }, { "epoch": 0.81, "learning_rate": 0.00019233643711260074, "loss": 1.5244, "step": 8169 }, { "epoch": 0.81, "learning_rate": 0.0001921482050903357, "loss": 1.4775, "step": 8170 }, { "epoch": 0.81, "learning_rate": 0.0001919600554305696, "loss": 1.4805, "step": 8171 }, { "epoch": 0.81, "learning_rate": 0.00019177198815248465, "loss": 1.3701, "step": 8172 }, { "epoch": 0.81, "learning_rate": 0.00019158400327525495, "loss": 1.5117, "step": 8173 }, { "epoch": 0.81, "learning_rate": 0.00019139610081804592, "loss": 1.4834, "step": 8174 }, { "epoch": 0.81, "learning_rate": 0.00019120828080001473, "loss": 1.3066, "step": 8175 }, { "epoch": 0.81, "learning_rate": 0.00019102054324031005, "loss": 1.4951, "step": 8176 }, { "epoch": 0.81, "learning_rate": 0.0001908328881580721, "loss": 1.4043, "step": 8177 }, { "epoch": 0.81, "learning_rate": 0.00019064531557243304, "loss": 1.4785, "step": 8178 }, { "epoch": 0.81, "learning_rate": 0.0001904578255025159, "loss": 1.3867, "step": 8179 }, { "epoch": 0.81, "learning_rate": 0.00019027041796743628, "loss": 1.2832, "step": 8180 }, { "epoch": 0.81, "learning_rate": 0.00019008309298630067, "loss": 1.3555, "step": 8181 }, { "epoch": 0.81, "learning_rate": 0.00018989585057820724, "loss": 1.46, "step": 8182 }, { "epoch": 0.81, "learning_rate": 0.00018970869076224595, "loss": 1.4346, "step": 8183 }, { "epoch": 0.81, "learning_rate": 0.00018952161355749808, "loss": 1.3433, "step": 8184 }, { "epoch": 0.81, "learning_rate": 0.0001893346189830366, "loss": 1.3838, "step": 8185 }, { "epoch": 0.81, "learning_rate": 0.00018914770705792628, "loss": 1.3994, "step": 8186 }, { "epoch": 0.81, "learning_rate": 0.00018896087780122297, "loss": 1.4268, "step": 8187 }, { "epoch": 0.81, "learning_rate": 0.0001887741312319745, "loss": 1.377, "step": 8188 }, { "epoch": 0.81, "learning_rate": 0.00018858746736922027, "loss": 1.3867, "step": 8189 }, { "epoch": 0.81, "learning_rate": 0.00018840088623199092, "loss": 1.3037, "step": 8190 }, { "epoch": 0.81, "learning_rate": 0.000188214387839309, "loss": 1.4229, "step": 8191 }, { "epoch": 0.81, "learning_rate": 0.00018802797221018842, "loss": 1.4834, "step": 8192 }, { "epoch": 0.81, "learning_rate": 0.00018784163936363463, "loss": 1.5244, "step": 8193 }, { "epoch": 0.81, "learning_rate": 0.000187655389318645, "loss": 1.4297, "step": 8194 }, { "epoch": 0.81, "learning_rate": 0.0001874692220942079, "loss": 1.2661, "step": 8195 }, { "epoch": 0.81, "learning_rate": 0.0001872831377093037, "loss": 1.3311, "step": 8196 }, { "epoch": 0.81, "learning_rate": 0.0001870971361829039, "loss": 1.4258, "step": 8197 }, { "epoch": 0.81, "learning_rate": 0.00018691121753397223, "loss": 1.4004, "step": 8198 }, { "epoch": 0.81, "learning_rate": 0.00018672538178146338, "loss": 1.5557, "step": 8199 }, { "epoch": 0.81, "learning_rate": 0.00018653962894432387, "loss": 1.4502, "step": 8200 }, { "epoch": 0.81, "learning_rate": 0.00018635395904149143, "loss": 1.4961, "step": 8201 }, { "epoch": 0.81, "learning_rate": 0.00018616837209189574, "loss": 1.4287, "step": 8202 }, { "epoch": 0.81, "learning_rate": 0.00018598286811445774, "loss": 1.3223, "step": 8203 }, { "epoch": 0.81, "learning_rate": 0.0001857974471280901, "loss": 1.2969, "step": 8204 }, { "epoch": 0.81, "learning_rate": 0.00018561210915169724, "loss": 1.4707, "step": 8205 }, { "epoch": 0.81, "learning_rate": 0.00018542685420417426, "loss": 1.3818, "step": 8206 }, { "epoch": 0.81, "learning_rate": 0.00018524168230440853, "loss": 1.332, "step": 8207 }, { "epoch": 0.81, "learning_rate": 0.00018505659347127902, "loss": 1.3877, "step": 8208 }, { "epoch": 0.81, "learning_rate": 0.00018487158772365577, "loss": 1.5088, "step": 8209 }, { "epoch": 0.81, "learning_rate": 0.00018468666508040065, "loss": 1.4414, "step": 8210 }, { "epoch": 0.81, "learning_rate": 0.00018450182556036676, "loss": 1.3271, "step": 8211 }, { "epoch": 0.81, "learning_rate": 0.00018431706918239942, "loss": 1.4814, "step": 8212 }, { "epoch": 0.81, "learning_rate": 0.0001841323959653347, "loss": 1.2539, "step": 8213 }, { "epoch": 0.81, "learning_rate": 0.00018394780592800064, "loss": 1.376, "step": 8214 }, { "epoch": 0.81, "learning_rate": 0.00018376329908921662, "loss": 1.4062, "step": 8215 }, { "epoch": 0.81, "learning_rate": 0.0001835788754677935, "loss": 1.4248, "step": 8216 }, { "epoch": 0.81, "learning_rate": 0.00018339453508253367, "loss": 1.5371, "step": 8217 }, { "epoch": 0.81, "learning_rate": 0.00018321027795223122, "loss": 1.3818, "step": 8218 }, { "epoch": 0.81, "learning_rate": 0.00018302610409567154, "loss": 1.4043, "step": 8219 }, { "epoch": 0.81, "learning_rate": 0.00018284201353163166, "loss": 1.4082, "step": 8220 }, { "epoch": 0.81, "learning_rate": 0.00018265800627888007, "loss": 1.2754, "step": 8221 }, { "epoch": 0.81, "learning_rate": 0.0001824740823561768, "loss": 1.4292, "step": 8222 }, { "epoch": 0.81, "learning_rate": 0.00018229024178227315, "loss": 1.4307, "step": 8223 }, { "epoch": 0.81, "learning_rate": 0.00018210648457591238, "loss": 1.375, "step": 8224 }, { "epoch": 0.81, "learning_rate": 0.00018192281075582895, "loss": 1.4961, "step": 8225 }, { "epoch": 0.81, "learning_rate": 0.00018173922034074876, "loss": 1.543, "step": 8226 }, { "epoch": 0.81, "learning_rate": 0.00018155571334938926, "loss": 1.5107, "step": 8227 }, { "epoch": 0.81, "learning_rate": 0.0001813722898004596, "loss": 1.4072, "step": 8228 }, { "epoch": 0.81, "learning_rate": 0.00018118894971265998, "loss": 1.4219, "step": 8229 }, { "epoch": 0.81, "learning_rate": 0.00018100569310468284, "loss": 1.4424, "step": 8230 }, { "epoch": 0.81, "learning_rate": 0.00018082251999521138, "loss": 1.4062, "step": 8231 }, { "epoch": 0.81, "learning_rate": 0.00018063943040292051, "loss": 1.2939, "step": 8232 }, { "epoch": 0.81, "learning_rate": 0.0001804564243464769, "loss": 1.5254, "step": 8233 }, { "epoch": 0.81, "learning_rate": 0.00018027350184453828, "loss": 1.334, "step": 8234 }, { "epoch": 0.81, "learning_rate": 0.00018009066291575405, "loss": 1.3271, "step": 8235 }, { "epoch": 0.81, "learning_rate": 0.00017990790757876508, "loss": 1.4204, "step": 8236 }, { "epoch": 0.81, "learning_rate": 0.0001797252358522038, "loss": 1.4492, "step": 8237 }, { "epoch": 0.81, "learning_rate": 0.00017954264775469408, "loss": 1.4961, "step": 8238 }, { "epoch": 0.81, "learning_rate": 0.00017936014330485118, "loss": 1.3525, "step": 8239 }, { "epoch": 0.81, "learning_rate": 0.00017917772252128183, "loss": 1.4199, "step": 8240 }, { "epoch": 0.81, "learning_rate": 0.0001789953854225843, "loss": 1.4189, "step": 8241 }, { "epoch": 0.81, "learning_rate": 0.0001788131320273484, "loss": 1.3916, "step": 8242 }, { "epoch": 0.81, "learning_rate": 0.0001786309623541553, "loss": 1.4482, "step": 8243 }, { "epoch": 0.81, "learning_rate": 0.0001784488764215776, "loss": 1.4492, "step": 8244 }, { "epoch": 0.81, "learning_rate": 0.00017826687424817933, "loss": 1.3232, "step": 8245 }, { "epoch": 0.81, "learning_rate": 0.00017808495585251626, "loss": 1.4727, "step": 8246 }, { "epoch": 0.81, "learning_rate": 0.000177903121253135, "loss": 1.4092, "step": 8247 }, { "epoch": 0.81, "learning_rate": 0.00017772137046857462, "loss": 1.5195, "step": 8248 }, { "epoch": 0.81, "learning_rate": 0.0001775397035173648, "loss": 1.334, "step": 8249 }, { "epoch": 0.81, "learning_rate": 0.00017735812041802702, "loss": 1.2939, "step": 8250 }, { "epoch": 0.81, "learning_rate": 0.00017717662118907408, "loss": 1.2295, "step": 8251 }, { "epoch": 0.81, "learning_rate": 0.0001769952058490101, "loss": 1.4326, "step": 8252 }, { "epoch": 0.81, "learning_rate": 0.000176813874416331, "loss": 1.2925, "step": 8253 }, { "epoch": 0.81, "learning_rate": 0.00017663262690952407, "loss": 1.3965, "step": 8254 }, { "epoch": 0.81, "learning_rate": 0.00017645146334706774, "loss": 1.4189, "step": 8255 }, { "epoch": 0.81, "learning_rate": 0.0001762703837474321, "loss": 1.4482, "step": 8256 }, { "epoch": 0.81, "learning_rate": 0.00017608938812907883, "loss": 1.5732, "step": 8257 }, { "epoch": 0.81, "learning_rate": 0.0001759084765104607, "loss": 1.3672, "step": 8258 }, { "epoch": 0.81, "learning_rate": 0.00017572764891002224, "loss": 1.3232, "step": 8259 }, { "epoch": 0.81, "learning_rate": 0.00017554690534619922, "loss": 1.4072, "step": 8260 }, { "epoch": 0.81, "learning_rate": 0.0001753662458374189, "loss": 1.3945, "step": 8261 }, { "epoch": 0.81, "learning_rate": 0.00017518567040209987, "loss": 1.4766, "step": 8262 }, { "epoch": 0.81, "learning_rate": 0.00017500517905865242, "loss": 1.3916, "step": 8263 }, { "epoch": 0.81, "learning_rate": 0.00017482477182547795, "loss": 1.377, "step": 8264 }, { "epoch": 0.81, "learning_rate": 0.0001746444487209695, "loss": 1.3643, "step": 8265 }, { "epoch": 0.81, "learning_rate": 0.0001744642097635115, "loss": 1.3389, "step": 8266 }, { "epoch": 0.81, "learning_rate": 0.0001742840549714796, "loss": 1.4004, "step": 8267 }, { "epoch": 0.82, "learning_rate": 0.00017410398436324114, "loss": 1.3682, "step": 8268 }, { "epoch": 0.82, "learning_rate": 0.00017392399795715475, "loss": 1.4443, "step": 8269 }, { "epoch": 0.82, "learning_rate": 0.0001737440957715706, "loss": 1.3105, "step": 8270 }, { "epoch": 0.82, "learning_rate": 0.00017356427782482998, "loss": 1.2578, "step": 8271 }, { "epoch": 0.82, "learning_rate": 0.00017338454413526582, "loss": 1.3604, "step": 8272 }, { "epoch": 0.82, "learning_rate": 0.0001732048947212025, "loss": 1.3164, "step": 8273 }, { "epoch": 0.82, "learning_rate": 0.00017302532960095562, "loss": 1.4258, "step": 8274 }, { "epoch": 0.82, "learning_rate": 0.00017284584879283238, "loss": 1.4023, "step": 8275 }, { "epoch": 0.82, "learning_rate": 0.00017266645231513124, "loss": 1.498, "step": 8276 }, { "epoch": 0.82, "learning_rate": 0.0001724871401861422, "loss": 1.458, "step": 8277 }, { "epoch": 0.82, "learning_rate": 0.00017230791242414645, "loss": 1.2627, "step": 8278 }, { "epoch": 0.82, "learning_rate": 0.00017212876904741659, "loss": 1.3438, "step": 8279 }, { "epoch": 0.82, "learning_rate": 0.0001719497100742171, "loss": 1.3984, "step": 8280 }, { "epoch": 0.82, "learning_rate": 0.00017177073552280333, "loss": 1.4131, "step": 8281 }, { "epoch": 0.82, "learning_rate": 0.00017159184541142215, "loss": 1.4355, "step": 8282 }, { "epoch": 0.82, "learning_rate": 0.0001714130397583118, "loss": 1.457, "step": 8283 }, { "epoch": 0.82, "learning_rate": 0.00017123431858170212, "loss": 1.3584, "step": 8284 }, { "epoch": 0.82, "learning_rate": 0.00017105568189981402, "loss": 1.3564, "step": 8285 }, { "epoch": 0.82, "learning_rate": 0.00017087712973086, "loss": 1.4512, "step": 8286 }, { "epoch": 0.82, "learning_rate": 0.00017069866209304385, "loss": 1.3467, "step": 8287 }, { "epoch": 0.82, "learning_rate": 0.0001705202790045609, "loss": 1.4463, "step": 8288 }, { "epoch": 0.82, "learning_rate": 0.00017034198048359773, "loss": 1.3896, "step": 8289 }, { "epoch": 0.82, "learning_rate": 0.00017016376654833222, "loss": 1.3809, "step": 8290 }, { "epoch": 0.82, "learning_rate": 0.00016998563721693383, "loss": 1.4609, "step": 8291 }, { "epoch": 0.82, "learning_rate": 0.00016980759250756329, "loss": 1.3418, "step": 8292 }, { "epoch": 0.82, "learning_rate": 0.00016962963243837248, "loss": 1.4365, "step": 8293 }, { "epoch": 0.82, "learning_rate": 0.00016945175702750526, "loss": 1.4482, "step": 8294 }, { "epoch": 0.82, "learning_rate": 0.0001692739662930961, "loss": 1.3179, "step": 8295 }, { "epoch": 0.82, "learning_rate": 0.00016909626025327142, "loss": 1.5811, "step": 8296 }, { "epoch": 0.82, "learning_rate": 0.00016891863892614845, "loss": 1.4014, "step": 8297 }, { "epoch": 0.82, "learning_rate": 0.00016874110232983664, "loss": 1.3301, "step": 8298 }, { "epoch": 0.82, "learning_rate": 0.00016856365048243616, "loss": 1.249, "step": 8299 }, { "epoch": 0.82, "learning_rate": 0.00016838628340203854, "loss": 1.3721, "step": 8300 }, { "epoch": 0.82, "learning_rate": 0.00016820900110672677, "loss": 1.3916, "step": 8301 }, { "epoch": 0.82, "learning_rate": 0.00016803180361457527, "loss": 1.4062, "step": 8302 }, { "epoch": 0.82, "learning_rate": 0.0001678546909436498, "loss": 1.3857, "step": 8303 }, { "epoch": 0.82, "learning_rate": 0.0001676776631120074, "loss": 1.4258, "step": 8304 }, { "epoch": 0.82, "learning_rate": 0.00016750072013769647, "loss": 1.4639, "step": 8305 }, { "epoch": 0.82, "learning_rate": 0.00016732386203875682, "loss": 1.457, "step": 8306 }, { "epoch": 0.82, "learning_rate": 0.00016714708883321962, "loss": 1.5156, "step": 8307 }, { "epoch": 0.82, "learning_rate": 0.00016697040053910718, "loss": 1.3779, "step": 8308 }, { "epoch": 0.82, "learning_rate": 0.00016679379717443355, "loss": 1.5664, "step": 8309 }, { "epoch": 0.82, "learning_rate": 0.00016661727875720366, "loss": 1.375, "step": 8310 }, { "epoch": 0.82, "learning_rate": 0.00016644084530541414, "loss": 1.2246, "step": 8311 }, { "epoch": 0.82, "learning_rate": 0.00016626449683705268, "loss": 1.4307, "step": 8312 }, { "epoch": 0.82, "learning_rate": 0.0001660882333700986, "loss": 1.2861, "step": 8313 }, { "epoch": 0.82, "learning_rate": 0.00016591205492252236, "loss": 1.4482, "step": 8314 }, { "epoch": 0.82, "learning_rate": 0.00016573596151228553, "loss": 1.5156, "step": 8315 }, { "epoch": 0.82, "learning_rate": 0.00016555995315734173, "loss": 1.4375, "step": 8316 }, { "epoch": 0.82, "learning_rate": 0.0001653840298756354, "loss": 1.4531, "step": 8317 }, { "epoch": 0.82, "learning_rate": 0.00016520819168510203, "loss": 1.4609, "step": 8318 }, { "epoch": 0.82, "learning_rate": 0.0001650324386036689, "loss": 1.3506, "step": 8319 }, { "epoch": 0.82, "learning_rate": 0.0001648567706492544, "loss": 1.4521, "step": 8320 }, { "epoch": 0.82, "learning_rate": 0.0001646811878397685, "loss": 1.4082, "step": 8321 }, { "epoch": 0.82, "learning_rate": 0.00016450569019311212, "loss": 1.3936, "step": 8322 }, { "epoch": 0.82, "learning_rate": 0.00016433027772717778, "loss": 1.4473, "step": 8323 }, { "epoch": 0.82, "learning_rate": 0.0001641549504598493, "loss": 1.3721, "step": 8324 }, { "epoch": 0.82, "learning_rate": 0.00016397970840900157, "loss": 1.3691, "step": 8325 }, { "epoch": 0.82, "learning_rate": 0.00016380455159250096, "loss": 1.415, "step": 8326 }, { "epoch": 0.82, "learning_rate": 0.00016362948002820532, "loss": 1.377, "step": 8327 }, { "epoch": 0.82, "learning_rate": 0.00016345449373396338, "loss": 1.3535, "step": 8328 }, { "epoch": 0.82, "learning_rate": 0.0001632795927276154, "loss": 1.334, "step": 8329 }, { "epoch": 0.82, "learning_rate": 0.00016310477702699333, "loss": 1.3574, "step": 8330 }, { "epoch": 0.82, "learning_rate": 0.00016293004664991983, "loss": 1.4404, "step": 8331 }, { "epoch": 0.82, "learning_rate": 0.0001627554016142092, "loss": 1.417, "step": 8332 }, { "epoch": 0.82, "learning_rate": 0.00016258084193766686, "loss": 1.4395, "step": 8333 }, { "epoch": 0.82, "learning_rate": 0.00016240636763808957, "loss": 1.4111, "step": 8334 }, { "epoch": 0.82, "learning_rate": 0.00016223197873326556, "loss": 1.4111, "step": 8335 }, { "epoch": 0.82, "learning_rate": 0.00016205767524097404, "loss": 1.3184, "step": 8336 }, { "epoch": 0.82, "learning_rate": 0.00016188345717898579, "loss": 1.4121, "step": 8337 }, { "epoch": 0.82, "learning_rate": 0.0001617093245650627, "loss": 1.4795, "step": 8338 }, { "epoch": 0.82, "learning_rate": 0.00016153527741695806, "loss": 1.376, "step": 8339 }, { "epoch": 0.82, "learning_rate": 0.0001613613157524164, "loss": 1.2446, "step": 8340 }, { "epoch": 0.82, "learning_rate": 0.0001611874395891736, "loss": 1.3701, "step": 8341 }, { "epoch": 0.82, "learning_rate": 0.00016101364894495663, "loss": 1.2886, "step": 8342 }, { "epoch": 0.82, "learning_rate": 0.00016083994383748413, "loss": 1.4824, "step": 8343 }, { "epoch": 0.82, "learning_rate": 0.00016066632428446547, "loss": 1.4258, "step": 8344 }, { "epoch": 0.82, "learning_rate": 0.00016049279030360176, "loss": 1.373, "step": 8345 }, { "epoch": 0.82, "learning_rate": 0.00016031934191258523, "loss": 1.4121, "step": 8346 }, { "epoch": 0.82, "learning_rate": 0.0001601459791290991, "loss": 1.3174, "step": 8347 }, { "epoch": 0.82, "learning_rate": 0.0001599727019708187, "loss": 1.3789, "step": 8348 }, { "epoch": 0.82, "learning_rate": 0.00015979951045540962, "loss": 1.4893, "step": 8349 }, { "epoch": 0.82, "learning_rate": 0.00015962640460052936, "loss": 1.4658, "step": 8350 }, { "epoch": 0.82, "learning_rate": 0.0001594533844238264, "loss": 1.3237, "step": 8351 }, { "epoch": 0.82, "learning_rate": 0.00015928044994294055, "loss": 1.3916, "step": 8352 }, { "epoch": 0.82, "learning_rate": 0.00015910760117550293, "loss": 1.4199, "step": 8353 }, { "epoch": 0.82, "learning_rate": 0.000158934838139136, "loss": 1.2471, "step": 8354 }, { "epoch": 0.82, "learning_rate": 0.00015876216085145334, "loss": 1.3848, "step": 8355 }, { "epoch": 0.82, "learning_rate": 0.00015858956933005974, "loss": 1.3369, "step": 8356 }, { "epoch": 0.82, "learning_rate": 0.00015841706359255147, "loss": 1.4395, "step": 8357 }, { "epoch": 0.82, "learning_rate": 0.00015824464365651582, "loss": 1.3984, "step": 8358 }, { "epoch": 0.82, "learning_rate": 0.00015807230953953144, "loss": 1.3857, "step": 8359 }, { "epoch": 0.82, "learning_rate": 0.00015790006125916822, "loss": 1.4521, "step": 8360 }, { "epoch": 0.82, "learning_rate": 0.00015772789883298732, "loss": 1.3555, "step": 8361 }, { "epoch": 0.82, "learning_rate": 0.00015755582227854116, "loss": 1.3643, "step": 8362 }, { "epoch": 0.82, "learning_rate": 0.00015738383161337333, "loss": 1.5527, "step": 8363 }, { "epoch": 0.82, "learning_rate": 0.0001572119268550186, "loss": 1.334, "step": 8364 }, { "epoch": 0.82, "learning_rate": 0.0001570401080210031, "loss": 1.4951, "step": 8365 }, { "epoch": 0.82, "learning_rate": 0.00015686837512884444, "loss": 1.3223, "step": 8366 }, { "epoch": 0.82, "learning_rate": 0.00015669672819605118, "loss": 1.4717, "step": 8367 }, { "epoch": 0.82, "learning_rate": 0.0001565251672401229, "loss": 1.4365, "step": 8368 }, { "epoch": 0.83, "learning_rate": 0.0001563536922785509, "loss": 1.4648, "step": 8369 }, { "epoch": 0.83, "learning_rate": 0.00015618230332881732, "loss": 1.3672, "step": 8370 }, { "epoch": 0.83, "learning_rate": 0.00015601100040839576, "loss": 1.3516, "step": 8371 }, { "epoch": 0.83, "learning_rate": 0.00015583978353475103, "loss": 1.416, "step": 8372 }, { "epoch": 0.83, "learning_rate": 0.00015566865272533925, "loss": 1.356, "step": 8373 }, { "epoch": 0.83, "learning_rate": 0.00015549760799760714, "loss": 1.4502, "step": 8374 }, { "epoch": 0.83, "learning_rate": 0.00015532664936899354, "loss": 1.3555, "step": 8375 }, { "epoch": 0.83, "learning_rate": 0.00015515577685692783, "loss": 1.4033, "step": 8376 }, { "epoch": 0.83, "learning_rate": 0.0001549849904788312, "loss": 1.3926, "step": 8377 }, { "epoch": 0.83, "learning_rate": 0.00015481429025211558, "loss": 1.3818, "step": 8378 }, { "epoch": 0.83, "learning_rate": 0.00015464367619418408, "loss": 1.416, "step": 8379 }, { "epoch": 0.83, "learning_rate": 0.00015447314832243164, "loss": 1.4775, "step": 8380 }, { "epoch": 0.83, "learning_rate": 0.00015430270665424384, "loss": 1.4209, "step": 8381 }, { "epoch": 0.83, "learning_rate": 0.00015413235120699753, "loss": 1.2642, "step": 8382 }, { "epoch": 0.83, "learning_rate": 0.00015396208199806107, "loss": 1.3345, "step": 8383 }, { "epoch": 0.83, "learning_rate": 0.00015379189904479363, "loss": 1.2246, "step": 8384 }, { "epoch": 0.83, "learning_rate": 0.00015362180236454582, "loss": 1.4209, "step": 8385 }, { "epoch": 0.83, "learning_rate": 0.00015345179197465952, "loss": 1.3955, "step": 8386 }, { "epoch": 0.83, "learning_rate": 0.00015328186789246757, "loss": 1.4551, "step": 8387 }, { "epoch": 0.83, "learning_rate": 0.00015311203013529428, "loss": 1.4326, "step": 8388 }, { "epoch": 0.83, "learning_rate": 0.00015294227872045507, "loss": 1.3301, "step": 8389 }, { "epoch": 0.83, "learning_rate": 0.0001527726136652563, "loss": 1.4092, "step": 8390 }, { "epoch": 0.83, "learning_rate": 0.000152603034986996, "loss": 1.3955, "step": 8391 }, { "epoch": 0.83, "learning_rate": 0.00015243354270296294, "loss": 1.3662, "step": 8392 }, { "epoch": 0.83, "learning_rate": 0.00015226413683043738, "loss": 1.3926, "step": 8393 }, { "epoch": 0.83, "learning_rate": 0.0001520948173866906, "loss": 1.4463, "step": 8394 }, { "epoch": 0.83, "learning_rate": 0.0001519255843889852, "loss": 1.3096, "step": 8395 }, { "epoch": 0.83, "learning_rate": 0.00015175643785457493, "loss": 1.3564, "step": 8396 }, { "epoch": 0.83, "learning_rate": 0.00015158737780070463, "loss": 1.5049, "step": 8397 }, { "epoch": 0.83, "learning_rate": 0.00015141840424461018, "loss": 1.5361, "step": 8398 }, { "epoch": 0.83, "learning_rate": 0.00015124951720351933, "loss": 1.4209, "step": 8399 }, { "epoch": 0.83, "learning_rate": 0.0001510807166946503, "loss": 1.3623, "step": 8400 }, { "epoch": 0.83, "learning_rate": 0.00015091200273521278, "loss": 1.4609, "step": 8401 }, { "epoch": 0.83, "learning_rate": 0.00015074337534240744, "loss": 1.4463, "step": 8402 }, { "epoch": 0.83, "learning_rate": 0.00015057483453342636, "loss": 1.4727, "step": 8403 }, { "epoch": 0.83, "learning_rate": 0.00015040638032545262, "loss": 1.4375, "step": 8404 }, { "epoch": 0.83, "learning_rate": 0.00015023801273566075, "loss": 1.4131, "step": 8405 }, { "epoch": 0.83, "learning_rate": 0.00015006973178121596, "loss": 1.2856, "step": 8406 }, { "epoch": 0.83, "learning_rate": 0.00014990153747927504, "loss": 1.4653, "step": 8407 }, { "epoch": 0.83, "learning_rate": 0.0001497334298469859, "loss": 1.3833, "step": 8408 }, { "epoch": 0.83, "learning_rate": 0.0001495654089014874, "loss": 1.3789, "step": 8409 }, { "epoch": 0.83, "learning_rate": 0.00014939747465990972, "loss": 1.2646, "step": 8410 }, { "epoch": 0.83, "learning_rate": 0.00014922962713937428, "loss": 1.4365, "step": 8411 }, { "epoch": 0.83, "learning_rate": 0.00014906186635699336, "loss": 1.4043, "step": 8412 }, { "epoch": 0.83, "learning_rate": 0.00014889419232987078, "loss": 1.3145, "step": 8413 }, { "epoch": 0.83, "learning_rate": 0.00014872660507510128, "loss": 1.3994, "step": 8414 }, { "epoch": 0.83, "learning_rate": 0.00014855910460977073, "loss": 1.4014, "step": 8415 }, { "epoch": 0.83, "learning_rate": 0.00014839169095095595, "loss": 1.4287, "step": 8416 }, { "epoch": 0.83, "learning_rate": 0.00014822436411572582, "loss": 1.3389, "step": 8417 }, { "epoch": 0.83, "learning_rate": 0.00014805712412113936, "loss": 1.2671, "step": 8418 }, { "epoch": 0.83, "learning_rate": 0.00014788997098424706, "loss": 1.3882, "step": 8419 }, { "epoch": 0.83, "learning_rate": 0.0001477229047220907, "loss": 1.3965, "step": 8420 }, { "epoch": 0.83, "learning_rate": 0.00014755592535170315, "loss": 1.3799, "step": 8421 }, { "epoch": 0.83, "learning_rate": 0.0001473890328901082, "loss": 1.4453, "step": 8422 }, { "epoch": 0.83, "learning_rate": 0.000147222227354321, "loss": 1.4102, "step": 8423 }, { "epoch": 0.83, "learning_rate": 0.00014705550876134788, "loss": 1.3193, "step": 8424 }, { "epoch": 0.83, "learning_rate": 0.0001468888771281862, "loss": 1.3418, "step": 8425 }, { "epoch": 0.83, "learning_rate": 0.00014672233247182444, "loss": 1.4141, "step": 8426 }, { "epoch": 0.83, "learning_rate": 0.00014655587480924225, "loss": 1.3428, "step": 8427 }, { "epoch": 0.83, "learning_rate": 0.00014638950415741026, "loss": 1.4707, "step": 8428 }, { "epoch": 0.83, "learning_rate": 0.00014622322053329062, "loss": 1.3281, "step": 8429 }, { "epoch": 0.83, "learning_rate": 0.00014605702395383623, "loss": 1.3506, "step": 8430 }, { "epoch": 0.83, "learning_rate": 0.00014589091443599123, "loss": 1.502, "step": 8431 }, { "epoch": 0.83, "learning_rate": 0.0001457248919966909, "loss": 1.3936, "step": 8432 }, { "epoch": 0.83, "learning_rate": 0.00014555895665286167, "loss": 1.2705, "step": 8433 }, { "epoch": 0.83, "learning_rate": 0.00014539310842142107, "loss": 1.4844, "step": 8434 }, { "epoch": 0.83, "learning_rate": 0.00014522734731927777, "loss": 1.4766, "step": 8435 }, { "epoch": 0.83, "learning_rate": 0.00014506167336333142, "loss": 1.2197, "step": 8436 }, { "epoch": 0.83, "learning_rate": 0.00014489608657047292, "loss": 1.4434, "step": 8437 }, { "epoch": 0.83, "learning_rate": 0.00014473058695758435, "loss": 1.3613, "step": 8438 }, { "epoch": 0.83, "learning_rate": 0.00014456517454153873, "loss": 1.499, "step": 8439 }, { "epoch": 0.83, "learning_rate": 0.00014439984933920026, "loss": 1.4434, "step": 8440 }, { "epoch": 0.83, "learning_rate": 0.00014423461136742433, "loss": 1.2769, "step": 8441 }, { "epoch": 0.83, "learning_rate": 0.00014406946064305726, "loss": 1.5176, "step": 8442 }, { "epoch": 0.83, "learning_rate": 0.0001439043971829366, "loss": 1.332, "step": 8443 }, { "epoch": 0.83, "learning_rate": 0.00014373942100389102, "loss": 1.3789, "step": 8444 }, { "epoch": 0.83, "learning_rate": 0.00014357453212274028, "loss": 1.4219, "step": 8445 }, { "epoch": 0.83, "learning_rate": 0.00014340973055629514, "loss": 1.4111, "step": 8446 }, { "epoch": 0.83, "learning_rate": 0.00014324501632135756, "loss": 1.3584, "step": 8447 }, { "epoch": 0.83, "learning_rate": 0.00014308038943472036, "loss": 1.3374, "step": 8448 }, { "epoch": 0.83, "learning_rate": 0.00014291584991316808, "loss": 1.4688, "step": 8449 }, { "epoch": 0.83, "learning_rate": 0.00014275139777347569, "loss": 1.6162, "step": 8450 }, { "epoch": 0.83, "learning_rate": 0.0001425870330324095, "loss": 1.3613, "step": 8451 }, { "epoch": 0.83, "learning_rate": 0.00014242275570672692, "loss": 1.3789, "step": 8452 }, { "epoch": 0.83, "learning_rate": 0.00014225856581317654, "loss": 1.3579, "step": 8453 }, { "epoch": 0.83, "learning_rate": 0.00014209446336849773, "loss": 1.3184, "step": 8454 }, { "epoch": 0.83, "learning_rate": 0.00014193044838942126, "loss": 1.3145, "step": 8455 }, { "epoch": 0.83, "learning_rate": 0.00014176652089266883, "loss": 1.3145, "step": 8456 }, { "epoch": 0.83, "learning_rate": 0.00014160268089495333, "loss": 1.4062, "step": 8457 }, { "epoch": 0.83, "learning_rate": 0.00014143892841297857, "loss": 1.377, "step": 8458 }, { "epoch": 0.83, "learning_rate": 0.00014127526346343955, "loss": 1.4121, "step": 8459 }, { "epoch": 0.83, "learning_rate": 0.00014111168606302228, "loss": 1.3066, "step": 8460 }, { "epoch": 0.83, "learning_rate": 0.00014094819622840395, "loss": 1.3608, "step": 8461 }, { "epoch": 0.83, "learning_rate": 0.00014078479397625266, "loss": 1.4092, "step": 8462 }, { "epoch": 0.83, "learning_rate": 0.00014062147932322776, "loss": 1.4678, "step": 8463 }, { "epoch": 0.83, "learning_rate": 0.00014045825228597953, "loss": 1.3662, "step": 8464 }, { "epoch": 0.83, "learning_rate": 0.0001402951128811495, "loss": 1.3525, "step": 8465 }, { "epoch": 0.83, "learning_rate": 0.0001401320611253698, "loss": 1.3809, "step": 8466 }, { "epoch": 0.83, "learning_rate": 0.0001399690970352644, "loss": 1.355, "step": 8467 }, { "epoch": 0.83, "learning_rate": 0.00013980622062744762, "loss": 1.3506, "step": 8468 }, { "epoch": 0.83, "learning_rate": 0.00013964343191852536, "loss": 1.4365, "step": 8469 }, { "epoch": 0.83, "learning_rate": 0.00013948073092509406, "loss": 1.3955, "step": 8470 }, { "epoch": 0.84, "learning_rate": 0.00013931811766374158, "loss": 1.46, "step": 8471 }, { "epoch": 0.84, "learning_rate": 0.00013915559215104678, "loss": 1.3994, "step": 8472 }, { "epoch": 0.84, "learning_rate": 0.0001389931544035795, "loss": 1.4219, "step": 8473 }, { "epoch": 0.84, "learning_rate": 0.00013883080443790076, "loss": 1.3838, "step": 8474 }, { "epoch": 0.84, "learning_rate": 0.00013866854227056246, "loss": 1.3008, "step": 8475 }, { "epoch": 0.84, "learning_rate": 0.00013850636791810767, "loss": 1.3906, "step": 8476 }, { "epoch": 0.84, "learning_rate": 0.00013834428139707046, "loss": 1.2969, "step": 8477 }, { "epoch": 0.84, "learning_rate": 0.00013818228272397592, "loss": 1.4023, "step": 8478 }, { "epoch": 0.84, "learning_rate": 0.00013802037191534022, "loss": 1.4033, "step": 8479 }, { "epoch": 0.84, "learning_rate": 0.0001378585489876706, "loss": 1.5039, "step": 8480 }, { "epoch": 0.84, "learning_rate": 0.00013769681395746526, "loss": 1.4141, "step": 8481 }, { "epoch": 0.84, "learning_rate": 0.00013753516684121358, "loss": 1.3691, "step": 8482 }, { "epoch": 0.84, "learning_rate": 0.00013737360765539585, "loss": 1.4492, "step": 8483 }, { "epoch": 0.84, "learning_rate": 0.00013721213641648312, "loss": 1.3867, "step": 8484 }, { "epoch": 0.84, "learning_rate": 0.0001370507531409384, "loss": 1.3965, "step": 8485 }, { "epoch": 0.84, "learning_rate": 0.0001368894578452149, "loss": 1.4717, "step": 8486 }, { "epoch": 0.84, "learning_rate": 0.00013672825054575677, "loss": 1.4941, "step": 8487 }, { "epoch": 0.84, "learning_rate": 0.00013656713125899977, "loss": 1.3896, "step": 8488 }, { "epoch": 0.84, "learning_rate": 0.00013640610000137032, "loss": 1.4487, "step": 8489 }, { "epoch": 0.84, "learning_rate": 0.00013624515678928596, "loss": 1.3818, "step": 8490 }, { "epoch": 0.84, "learning_rate": 0.0001360843016391552, "loss": 1.457, "step": 8491 }, { "epoch": 0.84, "learning_rate": 0.00013592353456737784, "loss": 1.4463, "step": 8492 }, { "epoch": 0.84, "learning_rate": 0.00013576285559034417, "loss": 1.3887, "step": 8493 }, { "epoch": 0.84, "learning_rate": 0.00013560226472443605, "loss": 1.3613, "step": 8494 }, { "epoch": 0.84, "learning_rate": 0.000135441761986026, "loss": 1.5312, "step": 8495 }, { "epoch": 0.84, "learning_rate": 0.0001352813473914778, "loss": 1.4014, "step": 8496 }, { "epoch": 0.84, "learning_rate": 0.00013512102095714597, "loss": 1.4131, "step": 8497 }, { "epoch": 0.84, "learning_rate": 0.00013496078269937596, "loss": 1.4277, "step": 8498 }, { "epoch": 0.84, "learning_rate": 0.00013480063263450503, "loss": 1.3467, "step": 8499 }, { "epoch": 0.84, "learning_rate": 0.00013464057077886062, "loss": 1.4287, "step": 8500 }, { "epoch": 0.84, "learning_rate": 0.00013448059714876137, "loss": 1.4023, "step": 8501 }, { "epoch": 0.84, "learning_rate": 0.00013432071176051695, "loss": 1.4824, "step": 8502 }, { "epoch": 0.84, "learning_rate": 0.00013416091463042834, "loss": 1.4473, "step": 8503 }, { "epoch": 0.84, "learning_rate": 0.0001340012057747869, "loss": 1.5225, "step": 8504 }, { "epoch": 0.84, "learning_rate": 0.00013384158520987556, "loss": 1.4238, "step": 8505 }, { "epoch": 0.84, "learning_rate": 0.000133682052951968, "loss": 1.3438, "step": 8506 }, { "epoch": 0.84, "learning_rate": 0.00013352260901732893, "loss": 1.4658, "step": 8507 }, { "epoch": 0.84, "learning_rate": 0.00013336325342221402, "loss": 1.4766, "step": 8508 }, { "epoch": 0.84, "learning_rate": 0.00013320398618287, "loss": 1.3779, "step": 8509 }, { "epoch": 0.84, "learning_rate": 0.0001330448073155345, "loss": 1.4746, "step": 8510 }, { "epoch": 0.84, "learning_rate": 0.00013288571683643636, "loss": 1.4795, "step": 8511 }, { "epoch": 0.84, "learning_rate": 0.00013272671476179497, "loss": 1.4639, "step": 8512 }, { "epoch": 0.84, "learning_rate": 0.0001325678011078213, "loss": 1.5283, "step": 8513 }, { "epoch": 0.84, "learning_rate": 0.00013240897589071677, "loss": 1.4219, "step": 8514 }, { "epoch": 0.84, "learning_rate": 0.0001322502391266741, "loss": 1.3232, "step": 8515 }, { "epoch": 0.84, "learning_rate": 0.00013209159083187673, "loss": 1.3828, "step": 8516 }, { "epoch": 0.84, "learning_rate": 0.00013193303102249955, "loss": 1.4424, "step": 8517 }, { "epoch": 0.84, "learning_rate": 0.000131774559714708, "loss": 1.4805, "step": 8518 }, { "epoch": 0.84, "learning_rate": 0.00013161617692465855, "loss": 1.4297, "step": 8519 }, { "epoch": 0.84, "learning_rate": 0.0001314578826684989, "loss": 1.3789, "step": 8520 }, { "epoch": 0.84, "learning_rate": 0.00013129967696236733, "loss": 1.3574, "step": 8521 }, { "epoch": 0.84, "learning_rate": 0.0001311415598223934, "loss": 1.4062, "step": 8522 }, { "epoch": 0.84, "learning_rate": 0.00013098353126469753, "loss": 1.3652, "step": 8523 }, { "epoch": 0.84, "learning_rate": 0.00013082559130539107, "loss": 1.4844, "step": 8524 }, { "epoch": 0.84, "learning_rate": 0.00013066773996057646, "loss": 1.3779, "step": 8525 }, { "epoch": 0.84, "learning_rate": 0.00013050997724634706, "loss": 1.3906, "step": 8526 }, { "epoch": 0.84, "learning_rate": 0.00013035230317878723, "loss": 1.4551, "step": 8527 }, { "epoch": 0.84, "learning_rate": 0.00013019471777397197, "loss": 1.373, "step": 8528 }, { "epoch": 0.84, "learning_rate": 0.00013003722104796777, "loss": 1.4453, "step": 8529 }, { "epoch": 0.84, "learning_rate": 0.0001298798130168317, "loss": 1.3975, "step": 8530 }, { "epoch": 0.84, "learning_rate": 0.00012972249369661194, "loss": 1.2871, "step": 8531 }, { "epoch": 0.84, "learning_rate": 0.0001295652631033476, "loss": 1.3516, "step": 8532 }, { "epoch": 0.84, "learning_rate": 0.00012940812125306856, "loss": 1.4658, "step": 8533 }, { "epoch": 0.84, "learning_rate": 0.00012925106816179588, "loss": 1.4199, "step": 8534 }, { "epoch": 0.84, "learning_rate": 0.00012909410384554177, "loss": 1.3438, "step": 8535 }, { "epoch": 0.84, "learning_rate": 0.00012893722832030884, "loss": 1.3369, "step": 8536 }, { "epoch": 0.84, "learning_rate": 0.00012878044160209113, "loss": 1.4463, "step": 8537 }, { "epoch": 0.84, "learning_rate": 0.00012862374370687347, "loss": 1.3809, "step": 8538 }, { "epoch": 0.84, "learning_rate": 0.0001284671346506313, "loss": 1.4668, "step": 8539 }, { "epoch": 0.84, "learning_rate": 0.00012831061444933155, "loss": 1.4888, "step": 8540 }, { "epoch": 0.84, "learning_rate": 0.00012815418311893178, "loss": 1.4307, "step": 8541 }, { "epoch": 0.84, "learning_rate": 0.0001279978406753809, "loss": 1.4404, "step": 8542 }, { "epoch": 0.84, "learning_rate": 0.00012784158713461768, "loss": 1.3857, "step": 8543 }, { "epoch": 0.84, "learning_rate": 0.000127685422512573, "loss": 1.3711, "step": 8544 }, { "epoch": 0.84, "learning_rate": 0.00012752934682516814, "loss": 1.3623, "step": 8545 }, { "epoch": 0.84, "learning_rate": 0.00012737336008831557, "loss": 1.3716, "step": 8546 }, { "epoch": 0.84, "learning_rate": 0.00012721746231791832, "loss": 1.3701, "step": 8547 }, { "epoch": 0.84, "learning_rate": 0.0001270616535298703, "loss": 1.2134, "step": 8548 }, { "epoch": 0.84, "learning_rate": 0.00012690593374005733, "loss": 1.5049, "step": 8549 }, { "epoch": 0.84, "learning_rate": 0.00012675030296435496, "loss": 1.4336, "step": 8550 }, { "epoch": 0.84, "learning_rate": 0.00012659476121863012, "loss": 1.4668, "step": 8551 }, { "epoch": 0.84, "learning_rate": 0.00012643930851874086, "loss": 1.4502, "step": 8552 }, { "epoch": 0.84, "learning_rate": 0.00012628394488053586, "loss": 1.4258, "step": 8553 }, { "epoch": 0.84, "learning_rate": 0.00012612867031985486, "loss": 1.4854, "step": 8554 }, { "epoch": 0.84, "learning_rate": 0.00012597348485252847, "loss": 1.4697, "step": 8555 }, { "epoch": 0.84, "learning_rate": 0.00012581838849437822, "loss": 1.3711, "step": 8556 }, { "epoch": 0.84, "learning_rate": 0.00012566338126121656, "loss": 1.2432, "step": 8557 }, { "epoch": 0.84, "learning_rate": 0.0001255084631688469, "loss": 1.3574, "step": 8558 }, { "epoch": 0.84, "learning_rate": 0.00012535363423306356, "loss": 1.4707, "step": 8559 }, { "epoch": 0.84, "learning_rate": 0.00012519889446965162, "loss": 1.3975, "step": 8560 }, { "epoch": 0.84, "learning_rate": 0.0001250442438943873, "loss": 1.335, "step": 8561 }, { "epoch": 0.84, "learning_rate": 0.00012488968252303756, "loss": 1.4629, "step": 8562 }, { "epoch": 0.84, "learning_rate": 0.00012473521037136026, "loss": 1.3564, "step": 8563 }, { "epoch": 0.84, "learning_rate": 0.00012458082745510436, "loss": 1.4609, "step": 8564 }, { "epoch": 0.84, "learning_rate": 0.00012442653379000944, "loss": 1.4258, "step": 8565 }, { "epoch": 0.84, "learning_rate": 0.000124272329391806, "loss": 1.4082, "step": 8566 }, { "epoch": 0.84, "learning_rate": 0.00012411821427621596, "loss": 1.2808, "step": 8567 }, { "epoch": 0.84, "learning_rate": 0.00012396418845895152, "loss": 1.3408, "step": 8568 }, { "epoch": 0.84, "learning_rate": 0.00012381025195571594, "loss": 1.417, "step": 8569 }, { "epoch": 0.84, "learning_rate": 0.00012365640478220353, "loss": 1.4316, "step": 8570 }, { "epoch": 0.84, "learning_rate": 0.0001235026469540994, "loss": 1.5352, "step": 8571 }, { "epoch": 0.85, "learning_rate": 0.00012334897848707938, "loss": 1.3779, "step": 8572 }, { "epoch": 0.85, "learning_rate": 0.00012319539939681056, "loss": 1.4131, "step": 8573 }, { "epoch": 0.85, "learning_rate": 0.00012304190969895058, "loss": 1.4453, "step": 8574 }, { "epoch": 0.85, "learning_rate": 0.0001228885094091481, "loss": 1.3691, "step": 8575 }, { "epoch": 0.85, "learning_rate": 0.00012273519854304272, "loss": 1.2471, "step": 8576 }, { "epoch": 0.85, "learning_rate": 0.00012258197711626485, "loss": 1.3721, "step": 8577 }, { "epoch": 0.85, "learning_rate": 0.00012242884514443575, "loss": 1.3955, "step": 8578 }, { "epoch": 0.85, "learning_rate": 0.00012227580264316773, "loss": 1.5078, "step": 8579 }, { "epoch": 0.85, "learning_rate": 0.00012212284962806363, "loss": 1.2764, "step": 8580 }, { "epoch": 0.85, "learning_rate": 0.00012196998611471766, "loss": 1.1934, "step": 8581 }, { "epoch": 0.85, "learning_rate": 0.00012181721211871433, "loss": 1.3936, "step": 8582 }, { "epoch": 0.85, "learning_rate": 0.00012166452765562963, "loss": 1.2349, "step": 8583 }, { "epoch": 0.85, "learning_rate": 0.00012151193274102967, "loss": 1.2881, "step": 8584 }, { "epoch": 0.85, "learning_rate": 0.00012135942739047246, "loss": 1.3398, "step": 8585 }, { "epoch": 0.85, "learning_rate": 0.00012120701161950609, "loss": 1.3896, "step": 8586 }, { "epoch": 0.85, "learning_rate": 0.0001210546854436696, "loss": 1.4443, "step": 8587 }, { "epoch": 0.85, "learning_rate": 0.00012090244887849311, "loss": 1.3594, "step": 8588 }, { "epoch": 0.85, "learning_rate": 0.00012075030193949743, "loss": 1.3936, "step": 8589 }, { "epoch": 0.85, "learning_rate": 0.00012059824464219448, "loss": 1.4443, "step": 8590 }, { "epoch": 0.85, "learning_rate": 0.00012044627700208677, "loss": 1.4414, "step": 8591 }, { "epoch": 0.85, "learning_rate": 0.00012029439903466777, "loss": 1.417, "step": 8592 }, { "epoch": 0.85, "learning_rate": 0.00012014261075542188, "loss": 1.3735, "step": 8593 }, { "epoch": 0.85, "learning_rate": 0.00011999091217982417, "loss": 1.25, "step": 8594 }, { "epoch": 0.85, "learning_rate": 0.00011983930332334092, "loss": 1.5244, "step": 8595 }, { "epoch": 0.85, "learning_rate": 0.00011968778420142879, "loss": 1.3682, "step": 8596 }, { "epoch": 0.85, "learning_rate": 0.00011953635482953562, "loss": 1.4746, "step": 8597 }, { "epoch": 0.85, "learning_rate": 0.00011938501522310008, "loss": 1.4004, "step": 8598 }, { "epoch": 0.85, "learning_rate": 0.00011923376539755148, "loss": 1.3027, "step": 8599 }, { "epoch": 0.85, "learning_rate": 0.00011908260536831018, "loss": 1.3467, "step": 8600 }, { "epoch": 0.85, "learning_rate": 0.00011893153515078736, "loss": 1.3887, "step": 8601 }, { "epoch": 0.85, "learning_rate": 0.00011878055476038497, "loss": 1.4111, "step": 8602 }, { "epoch": 0.85, "learning_rate": 0.00011862966421249588, "loss": 1.3081, "step": 8603 }, { "epoch": 0.85, "learning_rate": 0.0001184788635225037, "loss": 1.4043, "step": 8604 }, { "epoch": 0.85, "learning_rate": 0.0001183281527057829, "loss": 1.3882, "step": 8605 }, { "epoch": 0.85, "learning_rate": 0.00011817753177769896, "loss": 1.4736, "step": 8606 }, { "epoch": 0.85, "learning_rate": 0.00011802700075360795, "loss": 1.4678, "step": 8607 }, { "epoch": 0.85, "learning_rate": 0.0001178765596488569, "loss": 1.3652, "step": 8608 }, { "epoch": 0.85, "learning_rate": 0.00011772620847878357, "loss": 1.2852, "step": 8609 }, { "epoch": 0.85, "learning_rate": 0.00011757594725871679, "loss": 1.3457, "step": 8610 }, { "epoch": 0.85, "learning_rate": 0.00011742577600397608, "loss": 1.4941, "step": 8611 }, { "epoch": 0.85, "learning_rate": 0.00011727569472987154, "loss": 1.3623, "step": 8612 }, { "epoch": 0.85, "learning_rate": 0.00011712570345170459, "loss": 1.4863, "step": 8613 }, { "epoch": 0.85, "learning_rate": 0.00011697580218476711, "loss": 1.4072, "step": 8614 }, { "epoch": 0.85, "learning_rate": 0.00011682599094434188, "loss": 1.292, "step": 8615 }, { "epoch": 0.85, "learning_rate": 0.00011667626974570256, "loss": 1.4092, "step": 8616 }, { "epoch": 0.85, "learning_rate": 0.00011652663860411339, "loss": 1.3994, "step": 8617 }, { "epoch": 0.85, "learning_rate": 0.00011637709753482994, "loss": 1.4463, "step": 8618 }, { "epoch": 0.85, "learning_rate": 0.00011622764655309837, "loss": 1.4678, "step": 8619 }, { "epoch": 0.85, "learning_rate": 0.00011607828567415524, "loss": 1.1875, "step": 8620 }, { "epoch": 0.85, "learning_rate": 0.0001159290149132285, "loss": 1.3311, "step": 8621 }, { "epoch": 0.85, "learning_rate": 0.0001157798342855365, "loss": 1.3203, "step": 8622 }, { "epoch": 0.85, "learning_rate": 0.0001156307438062888, "loss": 1.3916, "step": 8623 }, { "epoch": 0.85, "learning_rate": 0.00011548174349068541, "loss": 1.2793, "step": 8624 }, { "epoch": 0.85, "learning_rate": 0.0001153328333539172, "loss": 1.4561, "step": 8625 }, { "epoch": 0.85, "learning_rate": 0.00011518401341116614, "loss": 1.3535, "step": 8626 }, { "epoch": 0.85, "learning_rate": 0.00011503528367760463, "loss": 1.3857, "step": 8627 }, { "epoch": 0.85, "learning_rate": 0.000114886644168396, "loss": 1.4355, "step": 8628 }, { "epoch": 0.85, "learning_rate": 0.00011473809489869457, "loss": 1.3779, "step": 8629 }, { "epoch": 0.85, "learning_rate": 0.00011458963588364524, "loss": 1.4697, "step": 8630 }, { "epoch": 0.85, "learning_rate": 0.00011444126713838365, "loss": 1.4932, "step": 8631 }, { "epoch": 0.85, "learning_rate": 0.00011429298867803661, "loss": 1.4951, "step": 8632 }, { "epoch": 0.85, "learning_rate": 0.00011414480051772136, "loss": 1.377, "step": 8633 }, { "epoch": 0.85, "learning_rate": 0.00011399670267254603, "loss": 1.4551, "step": 8634 }, { "epoch": 0.85, "learning_rate": 0.00011384869515760932, "loss": 1.4434, "step": 8635 }, { "epoch": 0.85, "learning_rate": 0.00011370077798800139, "loss": 1.5, "step": 8636 }, { "epoch": 0.85, "learning_rate": 0.00011355295117880271, "loss": 1.2778, "step": 8637 }, { "epoch": 0.85, "learning_rate": 0.00011340521474508436, "loss": 1.4033, "step": 8638 }, { "epoch": 0.85, "learning_rate": 0.00011325756870190862, "loss": 1.3916, "step": 8639 }, { "epoch": 0.85, "learning_rate": 0.00011311001306432822, "loss": 1.3242, "step": 8640 }, { "epoch": 0.85, "learning_rate": 0.00011296254784738702, "loss": 1.4238, "step": 8641 }, { "epoch": 0.85, "learning_rate": 0.00011281517306611921, "loss": 1.4053, "step": 8642 }, { "epoch": 0.85, "learning_rate": 0.00011266788873555022, "loss": 1.4243, "step": 8643 }, { "epoch": 0.85, "learning_rate": 0.00011252069487069604, "loss": 1.3809, "step": 8644 }, { "epoch": 0.85, "learning_rate": 0.00011237359148656334, "loss": 1.4541, "step": 8645 }, { "epoch": 0.85, "learning_rate": 0.00011222657859814977, "loss": 1.4219, "step": 8646 }, { "epoch": 0.85, "learning_rate": 0.00011207965622044358, "loss": 1.4805, "step": 8647 }, { "epoch": 0.85, "learning_rate": 0.000111932824368424, "loss": 1.3486, "step": 8648 }, { "epoch": 0.85, "learning_rate": 0.00011178608305706073, "loss": 1.5098, "step": 8649 }, { "epoch": 0.85, "learning_rate": 0.00011163943230131457, "loss": 1.4775, "step": 8650 }, { "epoch": 0.85, "learning_rate": 0.00011149287211613679, "loss": 1.4482, "step": 8651 }, { "epoch": 0.85, "learning_rate": 0.00011134640251646966, "loss": 1.2891, "step": 8652 }, { "epoch": 0.85, "learning_rate": 0.000111200023517246, "loss": 1.3115, "step": 8653 }, { "epoch": 0.85, "learning_rate": 0.00011105373513338968, "loss": 1.418, "step": 8654 }, { "epoch": 0.85, "learning_rate": 0.00011090753737981529, "loss": 1.4385, "step": 8655 }, { "epoch": 0.85, "learning_rate": 0.0001107614302714277, "loss": 1.3721, "step": 8656 }, { "epoch": 0.85, "learning_rate": 0.00011061541382312301, "loss": 1.4355, "step": 8657 }, { "epoch": 0.85, "learning_rate": 0.00011046948804978807, "loss": 1.4053, "step": 8658 }, { "epoch": 0.85, "learning_rate": 0.0001103236529663002, "loss": 1.4434, "step": 8659 }, { "epoch": 0.85, "learning_rate": 0.00011017790858752786, "loss": 1.3398, "step": 8660 }, { "epoch": 0.85, "learning_rate": 0.00011003225492832992, "loss": 1.4443, "step": 8661 }, { "epoch": 0.85, "learning_rate": 0.00010988669200355617, "loss": 1.4209, "step": 8662 }, { "epoch": 0.85, "learning_rate": 0.00010974121982804707, "loss": 1.5186, "step": 8663 }, { "epoch": 0.85, "learning_rate": 0.0001095958384166339, "loss": 1.3701, "step": 8664 }, { "epoch": 0.85, "learning_rate": 0.00010945054778413865, "loss": 1.3867, "step": 8665 }, { "epoch": 0.85, "learning_rate": 0.00010930534794537406, "loss": 1.4297, "step": 8666 }, { "epoch": 0.85, "learning_rate": 0.00010916023891514337, "loss": 1.4121, "step": 8667 }, { "epoch": 0.85, "learning_rate": 0.00010901522070824133, "loss": 1.4873, "step": 8668 }, { "epoch": 0.85, "learning_rate": 0.00010887029333945264, "loss": 1.3164, "step": 8669 }, { "epoch": 0.85, "learning_rate": 0.00010872545682355307, "loss": 1.4424, "step": 8670 }, { "epoch": 0.85, "learning_rate": 0.00010858071117530888, "loss": 1.3325, "step": 8671 }, { "epoch": 0.85, "learning_rate": 0.00010843605640947741, "loss": 1.3057, "step": 8672 }, { "epoch": 0.85, "learning_rate": 0.0001082914925408065, "loss": 1.2783, "step": 8673 }, { "epoch": 0.86, "learning_rate": 0.00010814701958403484, "loss": 1.4141, "step": 8674 }, { "epoch": 0.86, "learning_rate": 0.00010800263755389172, "loss": 1.4561, "step": 8675 }, { "epoch": 0.86, "learning_rate": 0.0001078583464650974, "loss": 1.2715, "step": 8676 }, { "epoch": 0.86, "learning_rate": 0.00010771414633236266, "loss": 1.2891, "step": 8677 }, { "epoch": 0.86, "learning_rate": 0.00010757003717038893, "loss": 1.3379, "step": 8678 }, { "epoch": 0.86, "learning_rate": 0.00010742601899386861, "loss": 1.4258, "step": 8679 }, { "epoch": 0.86, "learning_rate": 0.00010728209181748483, "loss": 1.3818, "step": 8680 }, { "epoch": 0.86, "learning_rate": 0.00010713825565591106, "loss": 1.4648, "step": 8681 }, { "epoch": 0.86, "learning_rate": 0.00010699451052381193, "loss": 1.4062, "step": 8682 }, { "epoch": 0.86, "learning_rate": 0.00010685085643584258, "loss": 1.418, "step": 8683 }, { "epoch": 0.86, "learning_rate": 0.00010670729340664886, "loss": 1.5605, "step": 8684 }, { "epoch": 0.86, "learning_rate": 0.00010656382145086729, "loss": 1.3457, "step": 8685 }, { "epoch": 0.86, "learning_rate": 0.00010642044058312539, "loss": 1.2485, "step": 8686 }, { "epoch": 0.86, "learning_rate": 0.00010627715081804124, "loss": 1.5518, "step": 8687 }, { "epoch": 0.86, "learning_rate": 0.00010613395217022337, "loss": 1.3115, "step": 8688 }, { "epoch": 0.86, "learning_rate": 0.00010599084465427145, "loss": 1.4736, "step": 8689 }, { "epoch": 0.86, "learning_rate": 0.00010584782828477546, "loss": 1.4238, "step": 8690 }, { "epoch": 0.86, "learning_rate": 0.0001057049030763163, "loss": 1.2568, "step": 8691 }, { "epoch": 0.86, "learning_rate": 0.00010556206904346566, "loss": 1.3623, "step": 8692 }, { "epoch": 0.86, "learning_rate": 0.00010541932620078576, "loss": 1.4023, "step": 8693 }, { "epoch": 0.86, "learning_rate": 0.00010527667456282964, "loss": 1.5, "step": 8694 }, { "epoch": 0.86, "learning_rate": 0.00010513411414414086, "loss": 1.293, "step": 8695 }, { "epoch": 0.86, "learning_rate": 0.00010499164495925396, "loss": 1.4004, "step": 8696 }, { "epoch": 0.86, "learning_rate": 0.00010484926702269393, "loss": 1.3125, "step": 8697 }, { "epoch": 0.86, "learning_rate": 0.00010470698034897653, "loss": 1.418, "step": 8698 }, { "epoch": 0.86, "learning_rate": 0.00010456478495260835, "loss": 1.3604, "step": 8699 }, { "epoch": 0.86, "learning_rate": 0.00010442268084808648, "loss": 1.3398, "step": 8700 }, { "epoch": 0.86, "learning_rate": 0.00010428066804989878, "loss": 1.3105, "step": 8701 }, { "epoch": 0.86, "learning_rate": 0.00010413874657252387, "loss": 1.4395, "step": 8702 }, { "epoch": 0.86, "learning_rate": 0.00010399691643043074, "loss": 1.3711, "step": 8703 }, { "epoch": 0.86, "learning_rate": 0.00010385517763807984, "loss": 1.4824, "step": 8704 }, { "epoch": 0.86, "learning_rate": 0.00010371353020992135, "loss": 1.2705, "step": 8705 }, { "epoch": 0.86, "learning_rate": 0.00010357197416039688, "loss": 1.5195, "step": 8706 }, { "epoch": 0.86, "learning_rate": 0.00010343050950393818, "loss": 1.502, "step": 8707 }, { "epoch": 0.86, "learning_rate": 0.00010328913625496805, "loss": 1.3311, "step": 8708 }, { "epoch": 0.86, "learning_rate": 0.00010314785442789987, "loss": 1.3906, "step": 8709 }, { "epoch": 0.86, "learning_rate": 0.00010300666403713754, "loss": 1.4248, "step": 8710 }, { "epoch": 0.86, "learning_rate": 0.00010286556509707612, "loss": 1.3662, "step": 8711 }, { "epoch": 0.86, "learning_rate": 0.00010272455762210042, "loss": 1.3555, "step": 8712 }, { "epoch": 0.86, "learning_rate": 0.00010258364162658684, "loss": 1.3467, "step": 8713 }, { "epoch": 0.86, "learning_rate": 0.0001024428171249021, "loss": 1.4658, "step": 8714 }, { "epoch": 0.86, "learning_rate": 0.00010230208413140362, "loss": 1.3369, "step": 8715 }, { "epoch": 0.86, "learning_rate": 0.00010216144266043936, "loss": 1.3867, "step": 8716 }, { "epoch": 0.86, "learning_rate": 0.00010202089272634807, "loss": 1.4287, "step": 8717 }, { "epoch": 0.86, "learning_rate": 0.0001018804343434594, "loss": 1.3413, "step": 8718 }, { "epoch": 0.86, "learning_rate": 0.00010174006752609322, "loss": 1.5361, "step": 8719 }, { "epoch": 0.86, "learning_rate": 0.00010159979228856042, "loss": 1.334, "step": 8720 }, { "epoch": 0.86, "learning_rate": 0.00010145960864516235, "loss": 1.4053, "step": 8721 }, { "epoch": 0.86, "learning_rate": 0.00010131951661019101, "loss": 1.3828, "step": 8722 }, { "epoch": 0.86, "learning_rate": 0.0001011795161979292, "loss": 1.4077, "step": 8723 }, { "epoch": 0.86, "learning_rate": 0.00010103960742265028, "loss": 1.4775, "step": 8724 }, { "epoch": 0.86, "learning_rate": 0.00010089979029861829, "loss": 1.4189, "step": 8725 }, { "epoch": 0.86, "learning_rate": 0.00010076006484008793, "loss": 1.3242, "step": 8726 }, { "epoch": 0.86, "learning_rate": 0.0001006204310613047, "loss": 1.2051, "step": 8727 }, { "epoch": 0.86, "learning_rate": 0.00010048088897650431, "loss": 1.6572, "step": 8728 }, { "epoch": 0.86, "learning_rate": 0.00010034143859991373, "loss": 1.4834, "step": 8729 }, { "epoch": 0.86, "learning_rate": 0.00010020207994575015, "loss": 1.5029, "step": 8730 }, { "epoch": 0.86, "learning_rate": 0.00010006281302822151, "loss": 1.5078, "step": 8731 }, { "epoch": 0.86, "learning_rate": 9.992363786152636e-05, "loss": 1.3252, "step": 8732 }, { "epoch": 0.86, "learning_rate": 9.978455445985412e-05, "loss": 1.4561, "step": 8733 }, { "epoch": 0.86, "learning_rate": 9.964556283738457e-05, "loss": 1.3345, "step": 8734 }, { "epoch": 0.86, "learning_rate": 9.950666300828803e-05, "loss": 1.3711, "step": 8735 }, { "epoch": 0.86, "learning_rate": 9.936785498672618e-05, "loss": 1.4795, "step": 8736 }, { "epoch": 0.86, "learning_rate": 9.922913878685058e-05, "loss": 1.3149, "step": 8737 }, { "epoch": 0.86, "learning_rate": 9.90905144228037e-05, "loss": 1.4336, "step": 8738 }, { "epoch": 0.86, "learning_rate": 9.895198190871868e-05, "loss": 1.228, "step": 8739 }, { "epoch": 0.86, "learning_rate": 9.881354125871922e-05, "loss": 1.4414, "step": 8740 }, { "epoch": 0.86, "learning_rate": 9.86751924869197e-05, "loss": 1.3071, "step": 8741 }, { "epoch": 0.86, "learning_rate": 9.853693560742505e-05, "loss": 1.3545, "step": 8742 }, { "epoch": 0.86, "learning_rate": 9.839877063433101e-05, "loss": 1.4326, "step": 8743 }, { "epoch": 0.86, "learning_rate": 9.826069758172373e-05, "loss": 1.2969, "step": 8744 }, { "epoch": 0.86, "learning_rate": 9.81227164636802e-05, "loss": 1.3184, "step": 8745 }, { "epoch": 0.86, "learning_rate": 9.798482729426794e-05, "loss": 1.3647, "step": 8746 }, { "epoch": 0.86, "learning_rate": 9.784703008754503e-05, "loss": 1.4414, "step": 8747 }, { "epoch": 0.86, "learning_rate": 9.770932485756023e-05, "loss": 1.3584, "step": 8748 }, { "epoch": 0.86, "learning_rate": 9.757171161835298e-05, "loss": 1.417, "step": 8749 }, { "epoch": 0.86, "learning_rate": 9.743419038395329e-05, "loss": 1.3691, "step": 8750 }, { "epoch": 0.86, "learning_rate": 9.729676116838182e-05, "loss": 1.4287, "step": 8751 }, { "epoch": 0.86, "learning_rate": 9.715942398564981e-05, "loss": 1.29, "step": 8752 }, { "epoch": 0.86, "learning_rate": 9.702217884975884e-05, "loss": 1.3506, "step": 8753 }, { "epoch": 0.86, "learning_rate": 9.688502577470182e-05, "loss": 1.3218, "step": 8754 }, { "epoch": 0.86, "learning_rate": 9.674796477446168e-05, "loss": 1.4697, "step": 8755 }, { "epoch": 0.86, "learning_rate": 9.661099586301214e-05, "loss": 1.5186, "step": 8756 }, { "epoch": 0.86, "learning_rate": 9.647411905431747e-05, "loss": 1.375, "step": 8757 }, { "epoch": 0.86, "learning_rate": 9.633733436233271e-05, "loss": 1.4102, "step": 8758 }, { "epoch": 0.86, "learning_rate": 9.62006418010033e-05, "loss": 1.4102, "step": 8759 }, { "epoch": 0.86, "learning_rate": 9.606404138426539e-05, "loss": 1.2476, "step": 8760 }, { "epoch": 0.86, "learning_rate": 9.592753312604574e-05, "loss": 1.3154, "step": 8761 }, { "epoch": 0.86, "learning_rate": 9.579111704026178e-05, "loss": 1.4492, "step": 8762 }, { "epoch": 0.86, "learning_rate": 9.565479314082137e-05, "loss": 1.3135, "step": 8763 }, { "epoch": 0.86, "learning_rate": 9.551856144162308e-05, "loss": 1.4229, "step": 8764 }, { "epoch": 0.86, "learning_rate": 9.538242195655622e-05, "loss": 1.3135, "step": 8765 }, { "epoch": 0.86, "learning_rate": 9.524637469950026e-05, "loss": 1.4131, "step": 8766 }, { "epoch": 0.86, "learning_rate": 9.511041968432588e-05, "loss": 1.2998, "step": 8767 }, { "epoch": 0.86, "learning_rate": 9.497455692489376e-05, "loss": 1.417, "step": 8768 }, { "epoch": 0.86, "learning_rate": 9.483878643505572e-05, "loss": 1.2871, "step": 8769 }, { "epoch": 0.86, "learning_rate": 9.470310822865358e-05, "loss": 1.4688, "step": 8770 }, { "epoch": 0.86, "learning_rate": 9.456752231952038e-05, "loss": 1.3174, "step": 8771 }, { "epoch": 0.86, "learning_rate": 9.443202872147915e-05, "loss": 1.4561, "step": 8772 }, { "epoch": 0.86, "learning_rate": 9.4296627448344e-05, "loss": 1.3779, "step": 8773 }, { "epoch": 0.86, "learning_rate": 9.41613185139194e-05, "loss": 1.3828, "step": 8774 }, { "epoch": 0.87, "learning_rate": 9.402610193200034e-05, "loss": 1.3555, "step": 8775 }, { "epoch": 0.87, "learning_rate": 9.389097771637268e-05, "loss": 1.3711, "step": 8776 }, { "epoch": 0.87, "learning_rate": 9.375594588081238e-05, "loss": 1.2681, "step": 8777 }, { "epoch": 0.87, "learning_rate": 9.362100643908656e-05, "loss": 1.4053, "step": 8778 }, { "epoch": 0.87, "learning_rate": 9.348615940495242e-05, "loss": 1.4072, "step": 8779 }, { "epoch": 0.87, "learning_rate": 9.33514047921581e-05, "loss": 1.3682, "step": 8780 }, { "epoch": 0.87, "learning_rate": 9.321674261444202e-05, "loss": 1.376, "step": 8781 }, { "epoch": 0.87, "learning_rate": 9.308217288553345e-05, "loss": 1.4326, "step": 8782 }, { "epoch": 0.87, "learning_rate": 9.294769561915206e-05, "loss": 1.4053, "step": 8783 }, { "epoch": 0.87, "learning_rate": 9.281331082900813e-05, "loss": 1.2441, "step": 8784 }, { "epoch": 0.87, "learning_rate": 9.267901852880234e-05, "loss": 1.4316, "step": 8785 }, { "epoch": 0.87, "learning_rate": 9.254481873222653e-05, "loss": 1.4375, "step": 8786 }, { "epoch": 0.87, "learning_rate": 9.241071145296253e-05, "loss": 1.3975, "step": 8787 }, { "epoch": 0.87, "learning_rate": 9.227669670468286e-05, "loss": 1.4199, "step": 8788 }, { "epoch": 0.87, "learning_rate": 9.214277450105057e-05, "loss": 1.5312, "step": 8789 }, { "epoch": 0.87, "learning_rate": 9.200894485571953e-05, "loss": 1.4111, "step": 8790 }, { "epoch": 0.87, "learning_rate": 9.187520778233394e-05, "loss": 1.3066, "step": 8791 }, { "epoch": 0.87, "learning_rate": 9.174156329452866e-05, "loss": 1.4414, "step": 8792 }, { "epoch": 0.87, "learning_rate": 9.1608011405929e-05, "loss": 1.5439, "step": 8793 }, { "epoch": 0.87, "learning_rate": 9.147455213015099e-05, "loss": 1.3296, "step": 8794 }, { "epoch": 0.87, "learning_rate": 9.134118548080107e-05, "loss": 1.313, "step": 8795 }, { "epoch": 0.87, "learning_rate": 9.120791147147633e-05, "loss": 1.4121, "step": 8796 }, { "epoch": 0.87, "learning_rate": 9.10747301157644e-05, "loss": 1.4141, "step": 8797 }, { "epoch": 0.87, "learning_rate": 9.094164142724337e-05, "loss": 1.3623, "step": 8798 }, { "epoch": 0.87, "learning_rate": 9.080864541948209e-05, "loss": 1.2432, "step": 8799 }, { "epoch": 0.87, "learning_rate": 9.06757421060398e-05, "loss": 1.3555, "step": 8800 }, { "epoch": 0.87, "learning_rate": 9.054293150046621e-05, "loss": 1.3486, "step": 8801 }, { "epoch": 0.87, "learning_rate": 9.041021361630175e-05, "loss": 1.4736, "step": 8802 }, { "epoch": 0.87, "learning_rate": 9.027758846707724e-05, "loss": 1.3779, "step": 8803 }, { "epoch": 0.87, "learning_rate": 9.014505606631451e-05, "loss": 1.4004, "step": 8804 }, { "epoch": 0.87, "learning_rate": 9.001261642752523e-05, "loss": 1.3682, "step": 8805 }, { "epoch": 0.87, "learning_rate": 8.988026956421214e-05, "loss": 1.3945, "step": 8806 }, { "epoch": 0.87, "learning_rate": 8.97480154898681e-05, "loss": 1.3716, "step": 8807 }, { "epoch": 0.87, "learning_rate": 8.961585421797691e-05, "loss": 1.4121, "step": 8808 }, { "epoch": 0.87, "learning_rate": 8.948378576201277e-05, "loss": 1.3369, "step": 8809 }, { "epoch": 0.87, "learning_rate": 8.935181013544036e-05, "loss": 1.4482, "step": 8810 }, { "epoch": 0.87, "learning_rate": 8.92199273517148e-05, "loss": 1.4434, "step": 8811 }, { "epoch": 0.87, "learning_rate": 8.908813742428201e-05, "loss": 1.3462, "step": 8812 }, { "epoch": 0.87, "learning_rate": 8.895644036657824e-05, "loss": 1.4053, "step": 8813 }, { "epoch": 0.87, "learning_rate": 8.88248361920303e-05, "loss": 1.3369, "step": 8814 }, { "epoch": 0.87, "learning_rate": 8.869332491405557e-05, "loss": 1.4229, "step": 8815 }, { "epoch": 0.87, "learning_rate": 8.85619065460621e-05, "loss": 1.4404, "step": 8816 }, { "epoch": 0.87, "learning_rate": 8.843058110144808e-05, "loss": 1.4229, "step": 8817 }, { "epoch": 0.87, "learning_rate": 8.829934859360266e-05, "loss": 1.2935, "step": 8818 }, { "epoch": 0.87, "learning_rate": 8.816820903590528e-05, "loss": 1.4014, "step": 8819 }, { "epoch": 0.87, "learning_rate": 8.803716244172588e-05, "loss": 1.4297, "step": 8820 }, { "epoch": 0.87, "learning_rate": 8.790620882442479e-05, "loss": 1.3672, "step": 8821 }, { "epoch": 0.87, "learning_rate": 8.777534819735355e-05, "loss": 1.4053, "step": 8822 }, { "epoch": 0.87, "learning_rate": 8.76445805738536e-05, "loss": 1.3418, "step": 8823 }, { "epoch": 0.87, "learning_rate": 8.751390596725673e-05, "loss": 1.3545, "step": 8824 }, { "epoch": 0.87, "learning_rate": 8.73833243908857e-05, "loss": 1.3315, "step": 8825 }, { "epoch": 0.87, "learning_rate": 8.725283585805355e-05, "loss": 1.4531, "step": 8826 }, { "epoch": 0.87, "learning_rate": 8.712244038206408e-05, "loss": 1.2964, "step": 8827 }, { "epoch": 0.87, "learning_rate": 8.69921379762113e-05, "loss": 1.4258, "step": 8828 }, { "epoch": 0.87, "learning_rate": 8.686192865377985e-05, "loss": 1.3604, "step": 8829 }, { "epoch": 0.87, "learning_rate": 8.673181242804507e-05, "loss": 1.4697, "step": 8830 }, { "epoch": 0.87, "learning_rate": 8.660178931227235e-05, "loss": 1.3291, "step": 8831 }, { "epoch": 0.87, "learning_rate": 8.647185931971813e-05, "loss": 1.4912, "step": 8832 }, { "epoch": 0.87, "learning_rate": 8.634202246362899e-05, "loss": 1.376, "step": 8833 }, { "epoch": 0.87, "learning_rate": 8.621227875724214e-05, "loss": 1.4004, "step": 8834 }, { "epoch": 0.87, "learning_rate": 8.608262821378521e-05, "loss": 1.3105, "step": 8835 }, { "epoch": 0.87, "learning_rate": 8.59530708464763e-05, "loss": 1.416, "step": 8836 }, { "epoch": 0.87, "learning_rate": 8.58236066685244e-05, "loss": 1.417, "step": 8837 }, { "epoch": 0.87, "learning_rate": 8.56942356931285e-05, "loss": 1.3984, "step": 8838 }, { "epoch": 0.87, "learning_rate": 8.556495793347852e-05, "loss": 1.4648, "step": 8839 }, { "epoch": 0.87, "learning_rate": 8.543577340275433e-05, "loss": 1.3789, "step": 8840 }, { "epoch": 0.87, "learning_rate": 8.530668211412674e-05, "loss": 1.4502, "step": 8841 }, { "epoch": 0.87, "learning_rate": 8.5177684080757e-05, "loss": 1.5195, "step": 8842 }, { "epoch": 0.87, "learning_rate": 8.50487793157968e-05, "loss": 1.2646, "step": 8843 }, { "epoch": 0.87, "learning_rate": 8.491996783238809e-05, "loss": 1.3936, "step": 8844 }, { "epoch": 0.87, "learning_rate": 8.479124964366381e-05, "loss": 1.4033, "step": 8845 }, { "epoch": 0.87, "learning_rate": 8.466262476274677e-05, "loss": 1.2383, "step": 8846 }, { "epoch": 0.87, "learning_rate": 8.453409320275096e-05, "loss": 1.3389, "step": 8847 }, { "epoch": 0.87, "learning_rate": 8.440565497678021e-05, "loss": 1.4014, "step": 8848 }, { "epoch": 0.87, "learning_rate": 8.427731009792917e-05, "loss": 1.3086, "step": 8849 }, { "epoch": 0.87, "learning_rate": 8.414905857928301e-05, "loss": 1.3564, "step": 8850 }, { "epoch": 0.87, "learning_rate": 8.40209004339173e-05, "loss": 1.3774, "step": 8851 }, { "epoch": 0.87, "learning_rate": 8.389283567489803e-05, "loss": 1.4639, "step": 8852 }, { "epoch": 0.87, "learning_rate": 8.376486431528162e-05, "loss": 1.4092, "step": 8853 }, { "epoch": 0.87, "learning_rate": 8.3636986368115e-05, "loss": 1.251, "step": 8854 }, { "epoch": 0.87, "learning_rate": 8.350920184643607e-05, "loss": 1.3643, "step": 8855 }, { "epoch": 0.87, "learning_rate": 8.338151076327249e-05, "loss": 1.3052, "step": 8856 }, { "epoch": 0.87, "learning_rate": 8.325391313164266e-05, "loss": 1.4248, "step": 8857 }, { "epoch": 0.87, "learning_rate": 8.312640896455558e-05, "loss": 1.4404, "step": 8858 }, { "epoch": 0.87, "learning_rate": 8.299899827501057e-05, "loss": 1.3672, "step": 8859 }, { "epoch": 0.87, "learning_rate": 8.287168107599741e-05, "loss": 1.3711, "step": 8860 }, { "epoch": 0.87, "learning_rate": 8.274445738049641e-05, "loss": 1.3223, "step": 8861 }, { "epoch": 0.87, "learning_rate": 8.261732720147852e-05, "loss": 1.3877, "step": 8862 }, { "epoch": 0.87, "learning_rate": 8.249029055190482e-05, "loss": 1.3916, "step": 8863 }, { "epoch": 0.87, "learning_rate": 8.236334744472696e-05, "loss": 1.4346, "step": 8864 }, { "epoch": 0.87, "learning_rate": 8.223649789288723e-05, "loss": 1.2334, "step": 8865 }, { "epoch": 0.87, "learning_rate": 8.21097419093182e-05, "loss": 1.4219, "step": 8866 }, { "epoch": 0.87, "learning_rate": 8.198307950694305e-05, "loss": 1.4248, "step": 8867 }, { "epoch": 0.87, "learning_rate": 8.185651069867517e-05, "loss": 1.3613, "step": 8868 }, { "epoch": 0.87, "learning_rate": 8.173003549741864e-05, "loss": 1.3364, "step": 8869 }, { "epoch": 0.87, "learning_rate": 8.160365391606794e-05, "loss": 1.3818, "step": 8870 }, { "epoch": 0.87, "learning_rate": 8.147736596750788e-05, "loss": 1.4473, "step": 8871 }, { "epoch": 0.87, "learning_rate": 8.135117166461392e-05, "loss": 1.3018, "step": 8872 }, { "epoch": 0.87, "learning_rate": 8.122507102025189e-05, "loss": 1.3213, "step": 8873 }, { "epoch": 0.87, "learning_rate": 8.109906404727819e-05, "loss": 1.332, "step": 8874 }, { "epoch": 0.87, "learning_rate": 8.097315075853928e-05, "loss": 1.4521, "step": 8875 }, { "epoch": 0.88, "learning_rate": 8.084733116687259e-05, "loss": 1.4727, "step": 8876 }, { "epoch": 0.88, "learning_rate": 8.072160528510553e-05, "loss": 1.2988, "step": 8877 }, { "epoch": 0.88, "learning_rate": 8.059597312605627e-05, "loss": 1.374, "step": 8878 }, { "epoch": 0.88, "learning_rate": 8.047043470253346e-05, "loss": 1.5371, "step": 8879 }, { "epoch": 0.88, "learning_rate": 8.034499002733575e-05, "loss": 1.4727, "step": 8880 }, { "epoch": 0.88, "learning_rate": 8.021963911325269e-05, "loss": 1.3975, "step": 8881 }, { "epoch": 0.88, "learning_rate": 8.009438197306407e-05, "loss": 1.4365, "step": 8882 }, { "epoch": 0.88, "learning_rate": 7.99692186195402e-05, "loss": 1.5088, "step": 8883 }, { "epoch": 0.88, "learning_rate": 7.98441490654418e-05, "loss": 1.5332, "step": 8884 }, { "epoch": 0.88, "learning_rate": 7.971917332352007e-05, "loss": 1.3994, "step": 8885 }, { "epoch": 0.88, "learning_rate": 7.95942914065163e-05, "loss": 1.3027, "step": 8886 }, { "epoch": 0.88, "learning_rate": 7.946950332716296e-05, "loss": 1.5342, "step": 8887 }, { "epoch": 0.88, "learning_rate": 7.934480909818231e-05, "loss": 1.4111, "step": 8888 }, { "epoch": 0.88, "learning_rate": 7.92202087322873e-05, "loss": 1.4951, "step": 8889 }, { "epoch": 0.88, "learning_rate": 7.90957022421811e-05, "loss": 1.3203, "step": 8890 }, { "epoch": 0.88, "learning_rate": 7.897128964055766e-05, "loss": 1.4834, "step": 8891 }, { "epoch": 0.88, "learning_rate": 7.884697094010085e-05, "loss": 1.3496, "step": 8892 }, { "epoch": 0.88, "learning_rate": 7.872274615348562e-05, "loss": 1.2754, "step": 8893 }, { "epoch": 0.88, "learning_rate": 7.859861529337675e-05, "loss": 1.4717, "step": 8894 }, { "epoch": 0.88, "learning_rate": 7.847457837242966e-05, "loss": 1.4668, "step": 8895 }, { "epoch": 0.88, "learning_rate": 7.835063540329046e-05, "loss": 1.4092, "step": 8896 }, { "epoch": 0.88, "learning_rate": 7.822678639859526e-05, "loss": 1.4443, "step": 8897 }, { "epoch": 0.88, "learning_rate": 7.810303137097085e-05, "loss": 1.3223, "step": 8898 }, { "epoch": 0.88, "learning_rate": 7.797937033303437e-05, "loss": 1.4961, "step": 8899 }, { "epoch": 0.88, "learning_rate": 7.785580329739339e-05, "loss": 1.4629, "step": 8900 }, { "epoch": 0.88, "learning_rate": 7.773233027664572e-05, "loss": 1.3896, "step": 8901 }, { "epoch": 0.88, "learning_rate": 7.760895128337986e-05, "loss": 1.3389, "step": 8902 }, { "epoch": 0.88, "learning_rate": 7.748566633017462e-05, "loss": 1.3535, "step": 8903 }, { "epoch": 0.88, "learning_rate": 7.736247542959895e-05, "loss": 1.3564, "step": 8904 }, { "epoch": 0.88, "learning_rate": 7.723937859421293e-05, "loss": 1.3926, "step": 8905 }, { "epoch": 0.88, "learning_rate": 7.711637583656639e-05, "loss": 1.4385, "step": 8906 }, { "epoch": 0.88, "learning_rate": 7.699346716919963e-05, "loss": 1.4395, "step": 8907 }, { "epoch": 0.88, "learning_rate": 7.687065260464365e-05, "loss": 1.4033, "step": 8908 }, { "epoch": 0.88, "learning_rate": 7.674793215541964e-05, "loss": 1.4717, "step": 8909 }, { "epoch": 0.88, "learning_rate": 7.662530583403937e-05, "loss": 1.4658, "step": 8910 }, { "epoch": 0.88, "learning_rate": 7.650277365300462e-05, "loss": 1.377, "step": 8911 }, { "epoch": 0.88, "learning_rate": 7.638033562480818e-05, "loss": 1.4307, "step": 8912 }, { "epoch": 0.88, "learning_rate": 7.62579917619326e-05, "loss": 1.501, "step": 8913 }, { "epoch": 0.88, "learning_rate": 7.613574207685148e-05, "loss": 1.3457, "step": 8914 }, { "epoch": 0.88, "learning_rate": 7.601358658202817e-05, "loss": 1.4033, "step": 8915 }, { "epoch": 0.88, "learning_rate": 7.589152528991682e-05, "loss": 1.5205, "step": 8916 }, { "epoch": 0.88, "learning_rate": 7.576955821296194e-05, "loss": 1.3701, "step": 8917 }, { "epoch": 0.88, "learning_rate": 7.564768536359845e-05, "loss": 1.3721, "step": 8918 }, { "epoch": 0.88, "learning_rate": 7.552590675425142e-05, "loss": 1.3086, "step": 8919 }, { "epoch": 0.88, "learning_rate": 7.540422239733647e-05, "loss": 1.3159, "step": 8920 }, { "epoch": 0.88, "learning_rate": 7.528263230525978e-05, "loss": 1.4658, "step": 8921 }, { "epoch": 0.88, "learning_rate": 7.516113649041755e-05, "loss": 1.4326, "step": 8922 }, { "epoch": 0.88, "learning_rate": 7.503973496519679e-05, "loss": 1.3159, "step": 8923 }, { "epoch": 0.88, "learning_rate": 7.491842774197466e-05, "loss": 1.4102, "step": 8924 }, { "epoch": 0.88, "learning_rate": 7.479721483311874e-05, "loss": 1.415, "step": 8925 }, { "epoch": 0.88, "learning_rate": 7.467609625098692e-05, "loss": 1.3945, "step": 8926 }, { "epoch": 0.88, "learning_rate": 7.455507200792755e-05, "loss": 1.3643, "step": 8927 }, { "epoch": 0.88, "learning_rate": 7.44341421162793e-05, "loss": 1.4102, "step": 8928 }, { "epoch": 0.88, "learning_rate": 7.431330658837133e-05, "loss": 1.4404, "step": 8929 }, { "epoch": 0.88, "learning_rate": 7.41925654365232e-05, "loss": 1.3174, "step": 8930 }, { "epoch": 0.88, "learning_rate": 7.407191867304464e-05, "loss": 1.376, "step": 8931 }, { "epoch": 0.88, "learning_rate": 7.395136631023591e-05, "loss": 1.4639, "step": 8932 }, { "epoch": 0.88, "learning_rate": 7.383090836038775e-05, "loss": 1.3623, "step": 8933 }, { "epoch": 0.88, "learning_rate": 7.3710544835781e-05, "loss": 1.4385, "step": 8934 }, { "epoch": 0.88, "learning_rate": 7.359027574868705e-05, "loss": 1.3652, "step": 8935 }, { "epoch": 0.88, "learning_rate": 7.347010111136765e-05, "loss": 1.4141, "step": 8936 }, { "epoch": 0.88, "learning_rate": 7.335002093607501e-05, "loss": 1.4248, "step": 8937 }, { "epoch": 0.88, "learning_rate": 7.323003523505134e-05, "loss": 1.3965, "step": 8938 }, { "epoch": 0.88, "learning_rate": 7.311014402052974e-05, "loss": 1.2988, "step": 8939 }, { "epoch": 0.88, "learning_rate": 7.29903473047333e-05, "loss": 1.3672, "step": 8940 }, { "epoch": 0.88, "learning_rate": 7.287064509987562e-05, "loss": 1.3945, "step": 8941 }, { "epoch": 0.88, "learning_rate": 7.275103741816069e-05, "loss": 1.4268, "step": 8942 }, { "epoch": 0.88, "learning_rate": 7.263152427178266e-05, "loss": 1.3965, "step": 8943 }, { "epoch": 0.88, "learning_rate": 7.251210567292631e-05, "loss": 1.4141, "step": 8944 }, { "epoch": 0.88, "learning_rate": 7.239278163376662e-05, "loss": 1.334, "step": 8945 }, { "epoch": 0.88, "learning_rate": 7.227355216646891e-05, "loss": 1.4443, "step": 8946 }, { "epoch": 0.88, "learning_rate": 7.215441728318916e-05, "loss": 1.4512, "step": 8947 }, { "epoch": 0.88, "learning_rate": 7.20353769960732e-05, "loss": 1.4268, "step": 8948 }, { "epoch": 0.88, "learning_rate": 7.191643131725755e-05, "loss": 1.3281, "step": 8949 }, { "epoch": 0.88, "learning_rate": 7.179758025886918e-05, "loss": 1.5488, "step": 8950 }, { "epoch": 0.88, "learning_rate": 7.167882383302505e-05, "loss": 1.3271, "step": 8951 }, { "epoch": 0.88, "learning_rate": 7.15601620518328e-05, "loss": 1.4795, "step": 8952 }, { "epoch": 0.88, "learning_rate": 7.144159492739023e-05, "loss": 1.3828, "step": 8953 }, { "epoch": 0.88, "learning_rate": 7.13231224717853e-05, "loss": 1.3613, "step": 8954 }, { "epoch": 0.88, "learning_rate": 7.120474469709715e-05, "loss": 1.4521, "step": 8955 }, { "epoch": 0.88, "learning_rate": 7.108646161539423e-05, "loss": 1.2549, "step": 8956 }, { "epoch": 0.88, "learning_rate": 7.096827323873611e-05, "loss": 1.3047, "step": 8957 }, { "epoch": 0.88, "learning_rate": 7.085017957917206e-05, "loss": 1.418, "step": 8958 }, { "epoch": 0.88, "learning_rate": 7.073218064874232e-05, "loss": 1.3906, "step": 8959 }, { "epoch": 0.88, "learning_rate": 7.061427645947693e-05, "loss": 1.3525, "step": 8960 }, { "epoch": 0.88, "learning_rate": 7.049646702339662e-05, "loss": 1.4385, "step": 8961 }, { "epoch": 0.88, "learning_rate": 7.037875235251245e-05, "loss": 1.4688, "step": 8962 }, { "epoch": 0.88, "learning_rate": 7.026113245882548e-05, "loss": 1.3896, "step": 8963 }, { "epoch": 0.88, "learning_rate": 7.014360735432767e-05, "loss": 1.439, "step": 8964 }, { "epoch": 0.88, "learning_rate": 7.002617705100067e-05, "loss": 1.3213, "step": 8965 }, { "epoch": 0.88, "learning_rate": 6.990884156081712e-05, "loss": 1.4619, "step": 8966 }, { "epoch": 0.88, "learning_rate": 6.979160089573933e-05, "loss": 1.397, "step": 8967 }, { "epoch": 0.88, "learning_rate": 6.967445506772052e-05, "loss": 1.2842, "step": 8968 }, { "epoch": 0.88, "learning_rate": 6.955740408870393e-05, "loss": 1.5088, "step": 8969 }, { "epoch": 0.88, "learning_rate": 6.944044797062309e-05, "loss": 1.3936, "step": 8970 }, { "epoch": 0.88, "learning_rate": 6.932358672540218e-05, "loss": 1.4062, "step": 8971 }, { "epoch": 0.88, "learning_rate": 6.920682036495518e-05, "loss": 1.3008, "step": 8972 }, { "epoch": 0.88, "learning_rate": 6.909014890118704e-05, "loss": 1.416, "step": 8973 }, { "epoch": 0.88, "learning_rate": 6.897357234599256e-05, "loss": 1.3965, "step": 8974 }, { "epoch": 0.88, "learning_rate": 6.885709071125712e-05, "loss": 1.4893, "step": 8975 }, { "epoch": 0.88, "learning_rate": 6.874070400885613e-05, "loss": 1.3906, "step": 8976 }, { "epoch": 0.88, "learning_rate": 6.862441225065564e-05, "loss": 1.46, "step": 8977 }, { "epoch": 0.89, "learning_rate": 6.850821544851183e-05, "loss": 1.3701, "step": 8978 }, { "epoch": 0.89, "learning_rate": 6.839211361427122e-05, "loss": 1.3228, "step": 8979 }, { "epoch": 0.89, "learning_rate": 6.82761067597707e-05, "loss": 1.3711, "step": 8980 }, { "epoch": 0.89, "learning_rate": 6.816019489683756e-05, "loss": 1.3623, "step": 8981 }, { "epoch": 0.89, "learning_rate": 6.804437803728914e-05, "loss": 1.4258, "step": 8982 }, { "epoch": 0.89, "learning_rate": 6.792865619293331e-05, "loss": 1.3989, "step": 8983 }, { "epoch": 0.89, "learning_rate": 6.781302937556833e-05, "loss": 1.5098, "step": 8984 }, { "epoch": 0.89, "learning_rate": 6.769749759698251e-05, "loss": 1.3691, "step": 8985 }, { "epoch": 0.89, "learning_rate": 6.758206086895458e-05, "loss": 1.3271, "step": 8986 }, { "epoch": 0.89, "learning_rate": 6.746671920325364e-05, "loss": 1.2715, "step": 8987 }, { "epoch": 0.89, "learning_rate": 6.73514726116391e-05, "loss": 1.377, "step": 8988 }, { "epoch": 0.89, "learning_rate": 6.723632110586053e-05, "loss": 1.2964, "step": 8989 }, { "epoch": 0.89, "learning_rate": 6.712126469765801e-05, "loss": 1.4209, "step": 8990 }, { "epoch": 0.89, "learning_rate": 6.700630339876179e-05, "loss": 1.4375, "step": 8991 }, { "epoch": 0.89, "learning_rate": 6.689143722089274e-05, "loss": 1.333, "step": 8992 }, { "epoch": 0.89, "learning_rate": 6.677666617576128e-05, "loss": 1.4688, "step": 8993 }, { "epoch": 0.89, "learning_rate": 6.66619902750688e-05, "loss": 1.2993, "step": 8994 }, { "epoch": 0.89, "learning_rate": 6.654740953050675e-05, "loss": 1.3799, "step": 8995 }, { "epoch": 0.89, "learning_rate": 6.643292395375711e-05, "loss": 1.46, "step": 8996 }, { "epoch": 0.89, "learning_rate": 6.631853355649175e-05, "loss": 1.373, "step": 8997 }, { "epoch": 0.89, "learning_rate": 6.620423835037315e-05, "loss": 1.3594, "step": 8998 }, { "epoch": 0.89, "learning_rate": 6.609003834705397e-05, "loss": 1.3809, "step": 8999 }, { "epoch": 0.89, "learning_rate": 6.597593355817722e-05, "loss": 1.459, "step": 9000 }, { "epoch": 0.89, "learning_rate": 6.586192399537605e-05, "loss": 1.4678, "step": 9001 }, { "epoch": 0.89, "learning_rate": 6.574800967027427e-05, "loss": 1.459, "step": 9002 }, { "epoch": 0.89, "learning_rate": 6.56341905944855e-05, "loss": 1.3643, "step": 9003 }, { "epoch": 0.89, "learning_rate": 6.552046677961376e-05, "loss": 1.3975, "step": 9004 }, { "epoch": 0.89, "learning_rate": 6.54068382372538e-05, "loss": 1.3838, "step": 9005 }, { "epoch": 0.89, "learning_rate": 6.529330497899022e-05, "loss": 1.376, "step": 9006 }, { "epoch": 0.89, "learning_rate": 6.5179867016398e-05, "loss": 1.3242, "step": 9007 }, { "epoch": 0.89, "learning_rate": 6.506652436104244e-05, "loss": 1.4473, "step": 9008 }, { "epoch": 0.89, "learning_rate": 6.495327702447907e-05, "loss": 1.3174, "step": 9009 }, { "epoch": 0.89, "learning_rate": 6.484012501825365e-05, "loss": 1.2988, "step": 9010 }, { "epoch": 0.89, "learning_rate": 6.472706835390251e-05, "loss": 1.459, "step": 9011 }, { "epoch": 0.89, "learning_rate": 6.46141070429519e-05, "loss": 1.3066, "step": 9012 }, { "epoch": 0.89, "learning_rate": 6.450124109691858e-05, "loss": 1.4014, "step": 9013 }, { "epoch": 0.89, "learning_rate": 6.438847052730956e-05, "loss": 1.2817, "step": 9014 }, { "epoch": 0.89, "learning_rate": 6.427579534562189e-05, "loss": 1.3047, "step": 9015 }, { "epoch": 0.89, "learning_rate": 6.416321556334337e-05, "loss": 1.416, "step": 9016 }, { "epoch": 0.89, "learning_rate": 6.405073119195148e-05, "loss": 1.4883, "step": 9017 }, { "epoch": 0.89, "learning_rate": 6.39383422429145e-05, "loss": 1.4209, "step": 9018 }, { "epoch": 0.89, "learning_rate": 6.382604872769071e-05, "loss": 1.4512, "step": 9019 }, { "epoch": 0.89, "learning_rate": 6.371385065772862e-05, "loss": 1.3252, "step": 9020 }, { "epoch": 0.89, "learning_rate": 6.360174804446717e-05, "loss": 1.4023, "step": 9021 }, { "epoch": 0.89, "learning_rate": 6.348974089933546e-05, "loss": 1.3208, "step": 9022 }, { "epoch": 0.89, "learning_rate": 6.337782923375302e-05, "loss": 1.3921, "step": 9023 }, { "epoch": 0.89, "learning_rate": 6.326601305912949e-05, "loss": 1.3262, "step": 9024 }, { "epoch": 0.89, "learning_rate": 6.315429238686465e-05, "loss": 1.3887, "step": 9025 }, { "epoch": 0.89, "learning_rate": 6.304266722834894e-05, "loss": 1.4307, "step": 9026 }, { "epoch": 0.89, "learning_rate": 6.29311375949626e-05, "loss": 1.3916, "step": 9027 }, { "epoch": 0.89, "learning_rate": 6.281970349807653e-05, "loss": 1.4014, "step": 9028 }, { "epoch": 0.89, "learning_rate": 6.270836494905163e-05, "loss": 1.3896, "step": 9029 }, { "epoch": 0.89, "learning_rate": 6.259712195923905e-05, "loss": 1.3496, "step": 9030 }, { "epoch": 0.89, "learning_rate": 6.24859745399805e-05, "loss": 1.5088, "step": 9031 }, { "epoch": 0.89, "learning_rate": 6.237492270260758e-05, "loss": 1.4912, "step": 9032 }, { "epoch": 0.89, "learning_rate": 6.226396645844234e-05, "loss": 1.4238, "step": 9033 }, { "epoch": 0.89, "learning_rate": 6.215310581879707e-05, "loss": 1.4814, "step": 9034 }, { "epoch": 0.89, "learning_rate": 6.204234079497418e-05, "loss": 1.4824, "step": 9035 }, { "epoch": 0.89, "learning_rate": 6.193167139826661e-05, "loss": 1.3301, "step": 9036 }, { "epoch": 0.89, "learning_rate": 6.182109763995725e-05, "loss": 1.3613, "step": 9037 }, { "epoch": 0.89, "learning_rate": 6.17106195313194e-05, "loss": 1.4824, "step": 9038 }, { "epoch": 0.89, "learning_rate": 6.160023708361662e-05, "loss": 1.3301, "step": 9039 }, { "epoch": 0.89, "learning_rate": 6.148995030810245e-05, "loss": 1.3633, "step": 9040 }, { "epoch": 0.89, "learning_rate": 6.137975921602124e-05, "loss": 1.3193, "step": 9041 }, { "epoch": 0.89, "learning_rate": 6.126966381860722e-05, "loss": 1.3057, "step": 9042 }, { "epoch": 0.89, "learning_rate": 6.115966412708463e-05, "loss": 1.4717, "step": 9043 }, { "epoch": 0.89, "learning_rate": 6.10497601526685e-05, "loss": 1.4326, "step": 9044 }, { "epoch": 0.89, "learning_rate": 6.093995190656354e-05, "loss": 1.3516, "step": 9045 }, { "epoch": 0.89, "learning_rate": 6.083023939996513e-05, "loss": 1.3389, "step": 9046 }, { "epoch": 0.89, "learning_rate": 6.072062264405876e-05, "loss": 1.3384, "step": 9047 }, { "epoch": 0.89, "learning_rate": 6.0611101650020154e-05, "loss": 1.4629, "step": 9048 }, { "epoch": 0.89, "learning_rate": 6.050167642901505e-05, "loss": 1.4004, "step": 9049 }, { "epoch": 0.89, "learning_rate": 6.039234699219975e-05, "loss": 1.3232, "step": 9050 }, { "epoch": 0.89, "learning_rate": 6.0283113350720674e-05, "loss": 1.3701, "step": 9051 }, { "epoch": 0.89, "learning_rate": 6.017397551571435e-05, "loss": 1.4609, "step": 9052 }, { "epoch": 0.89, "learning_rate": 6.0064933498307774e-05, "loss": 1.3965, "step": 9053 }, { "epoch": 0.89, "learning_rate": 5.9955987309618045e-05, "loss": 1.2607, "step": 9054 }, { "epoch": 0.89, "learning_rate": 5.984713696075239e-05, "loss": 1.374, "step": 9055 }, { "epoch": 0.89, "learning_rate": 5.97383824628085e-05, "loss": 1.5195, "step": 9056 }, { "epoch": 0.89, "learning_rate": 5.962972382687415e-05, "loss": 1.4629, "step": 9057 }, { "epoch": 0.89, "learning_rate": 5.9521161064027385e-05, "loss": 1.4717, "step": 9058 }, { "epoch": 0.89, "learning_rate": 5.941269418533635e-05, "loss": 1.4014, "step": 9059 }, { "epoch": 0.89, "learning_rate": 5.930432320185952e-05, "loss": 1.4004, "step": 9060 }, { "epoch": 0.89, "learning_rate": 5.919604812464574e-05, "loss": 1.4375, "step": 9061 }, { "epoch": 0.89, "learning_rate": 5.9087868964733725e-05, "loss": 1.2427, "step": 9062 }, { "epoch": 0.89, "learning_rate": 5.897978573315266e-05, "loss": 1.377, "step": 9063 }, { "epoch": 0.89, "learning_rate": 5.8871798440922054e-05, "loss": 1.4648, "step": 9064 }, { "epoch": 0.89, "learning_rate": 5.8763907099051214e-05, "loss": 1.4863, "step": 9065 }, { "epoch": 0.89, "learning_rate": 5.865611171854013e-05, "loss": 1.4814, "step": 9066 }, { "epoch": 0.89, "learning_rate": 5.8548412310378774e-05, "loss": 1.3613, "step": 9067 }, { "epoch": 0.89, "learning_rate": 5.844080888554726e-05, "loss": 1.4072, "step": 9068 }, { "epoch": 0.89, "learning_rate": 5.8333301455016255e-05, "loss": 1.2998, "step": 9069 }, { "epoch": 0.89, "learning_rate": 5.822589002974621e-05, "loss": 1.4141, "step": 9070 }, { "epoch": 0.89, "learning_rate": 5.811857462068793e-05, "loss": 1.417, "step": 9071 }, { "epoch": 0.89, "learning_rate": 5.801135523878276e-05, "loss": 1.3633, "step": 9072 }, { "epoch": 0.89, "learning_rate": 5.7904231894961526e-05, "loss": 1.3291, "step": 9073 }, { "epoch": 0.89, "learning_rate": 5.7797204600146256e-05, "loss": 1.2598, "step": 9074 }, { "epoch": 0.89, "learning_rate": 5.7690273365248344e-05, "loss": 1.4414, "step": 9075 }, { "epoch": 0.89, "learning_rate": 5.758343820116985e-05, "loss": 1.4746, "step": 9076 }, { "epoch": 0.89, "learning_rate": 5.747669911880271e-05, "loss": 1.4238, "step": 9077 }, { "epoch": 0.89, "learning_rate": 5.737005612902946e-05, "loss": 1.3213, "step": 9078 }, { "epoch": 0.9, "learning_rate": 5.726350924272228e-05, "loss": 1.3208, "step": 9079 }, { "epoch": 0.9, "learning_rate": 5.715705847074426e-05, "loss": 1.3174, "step": 9080 }, { "epoch": 0.9, "learning_rate": 5.705070382394806e-05, "loss": 1.4912, "step": 9081 }, { "epoch": 0.9, "learning_rate": 5.6944445313176886e-05, "loss": 1.3398, "step": 9082 }, { "epoch": 0.9, "learning_rate": 5.6838282949264076e-05, "loss": 1.4072, "step": 9083 }, { "epoch": 0.9, "learning_rate": 5.6732216743033194e-05, "loss": 1.3496, "step": 9084 }, { "epoch": 0.9, "learning_rate": 5.6626246705297815e-05, "loss": 1.2695, "step": 9085 }, { "epoch": 0.9, "learning_rate": 5.6520372846861954e-05, "loss": 1.29, "step": 9086 }, { "epoch": 0.9, "learning_rate": 5.6414595178519654e-05, "loss": 1.3672, "step": 9087 }, { "epoch": 0.9, "learning_rate": 5.630891371105529e-05, "loss": 1.3135, "step": 9088 }, { "epoch": 0.9, "learning_rate": 5.620332845524334e-05, "loss": 1.3984, "step": 9089 }, { "epoch": 0.9, "learning_rate": 5.609783942184843e-05, "loss": 1.3848, "step": 9090 }, { "epoch": 0.9, "learning_rate": 5.5992446621625285e-05, "loss": 1.3828, "step": 9091 }, { "epoch": 0.9, "learning_rate": 5.5887150065319195e-05, "loss": 1.5049, "step": 9092 }, { "epoch": 0.9, "learning_rate": 5.578194976366546e-05, "loss": 1.4531, "step": 9093 }, { "epoch": 0.9, "learning_rate": 5.567684572738929e-05, "loss": 1.3838, "step": 9094 }, { "epoch": 0.9, "learning_rate": 5.557183796720644e-05, "loss": 1.3457, "step": 9095 }, { "epoch": 0.9, "learning_rate": 5.5466926493822675e-05, "loss": 1.4756, "step": 9096 }, { "epoch": 0.9, "learning_rate": 5.536211131793389e-05, "loss": 1.3232, "step": 9097 }, { "epoch": 0.9, "learning_rate": 5.5257392450226426e-05, "loss": 1.376, "step": 9098 }, { "epoch": 0.9, "learning_rate": 5.515276990137663e-05, "loss": 1.2964, "step": 9099 }, { "epoch": 0.9, "learning_rate": 5.5048243682050855e-05, "loss": 1.4736, "step": 9100 }, { "epoch": 0.9, "learning_rate": 5.4943813802905915e-05, "loss": 1.4697, "step": 9101 }, { "epoch": 0.9, "learning_rate": 5.483948027458863e-05, "loss": 1.3916, "step": 9102 }, { "epoch": 0.9, "learning_rate": 5.473524310773614e-05, "loss": 1.457, "step": 9103 }, { "epoch": 0.9, "learning_rate": 5.463110231297585e-05, "loss": 1.2773, "step": 9104 }, { "epoch": 0.9, "learning_rate": 5.45270579009246e-05, "loss": 1.4502, "step": 9105 }, { "epoch": 0.9, "learning_rate": 5.442310988219057e-05, "loss": 1.54, "step": 9106 }, { "epoch": 0.9, "learning_rate": 5.431925826737138e-05, "loss": 1.3486, "step": 9107 }, { "epoch": 0.9, "learning_rate": 5.4215503067054805e-05, "loss": 1.3584, "step": 9108 }, { "epoch": 0.9, "learning_rate": 5.411184429181903e-05, "loss": 1.4307, "step": 9109 }, { "epoch": 0.9, "learning_rate": 5.400828195223229e-05, "loss": 1.4072, "step": 9110 }, { "epoch": 0.9, "learning_rate": 5.390481605885311e-05, "loss": 1.3608, "step": 9111 }, { "epoch": 0.9, "learning_rate": 5.3801446622229965e-05, "loss": 1.4067, "step": 9112 }, { "epoch": 0.9, "learning_rate": 5.369817365290175e-05, "loss": 1.3359, "step": 9113 }, { "epoch": 0.9, "learning_rate": 5.3594997161397374e-05, "loss": 1.3662, "step": 9114 }, { "epoch": 0.9, "learning_rate": 5.349191715823587e-05, "loss": 1.4512, "step": 9115 }, { "epoch": 0.9, "learning_rate": 5.338893365392661e-05, "loss": 1.4824, "step": 9116 }, { "epoch": 0.9, "learning_rate": 5.328604665896897e-05, "loss": 1.4785, "step": 9117 }, { "epoch": 0.9, "learning_rate": 5.318325618385245e-05, "loss": 1.4639, "step": 9118 }, { "epoch": 0.9, "learning_rate": 5.3080562239056886e-05, "loss": 1.543, "step": 9119 }, { "epoch": 0.9, "learning_rate": 5.297796483505213e-05, "loss": 1.4844, "step": 9120 }, { "epoch": 0.9, "learning_rate": 5.287546398229826e-05, "loss": 1.4531, "step": 9121 }, { "epoch": 0.9, "learning_rate": 5.2773059691245576e-05, "loss": 1.4131, "step": 9122 }, { "epoch": 0.9, "learning_rate": 5.2670751972334175e-05, "loss": 1.459, "step": 9123 }, { "epoch": 0.9, "learning_rate": 5.2568540835994935e-05, "loss": 1.2656, "step": 9124 }, { "epoch": 0.9, "learning_rate": 5.246642629264842e-05, "loss": 1.1826, "step": 9125 }, { "epoch": 0.9, "learning_rate": 5.23644083527054e-05, "loss": 1.3398, "step": 9126 }, { "epoch": 0.9, "learning_rate": 5.2262487026566884e-05, "loss": 1.251, "step": 9127 }, { "epoch": 0.9, "learning_rate": 5.216066232462402e-05, "loss": 1.3691, "step": 9128 }, { "epoch": 0.9, "learning_rate": 5.205893425725816e-05, "loss": 1.4102, "step": 9129 }, { "epoch": 0.9, "learning_rate": 5.195730283484057e-05, "loss": 1.2812, "step": 9130 }, { "epoch": 0.9, "learning_rate": 5.1855768067732846e-05, "loss": 1.3989, "step": 9131 }, { "epoch": 0.9, "learning_rate": 5.175432996628682e-05, "loss": 1.311, "step": 9132 }, { "epoch": 0.9, "learning_rate": 5.165298854084433e-05, "loss": 1.3223, "step": 9133 }, { "epoch": 0.9, "learning_rate": 5.1551743801737326e-05, "loss": 1.3457, "step": 9134 }, { "epoch": 0.9, "learning_rate": 5.145059575928801e-05, "loss": 1.3955, "step": 9135 }, { "epoch": 0.9, "learning_rate": 5.134954442380868e-05, "loss": 1.2456, "step": 9136 }, { "epoch": 0.9, "learning_rate": 5.1248589805601766e-05, "loss": 1.4443, "step": 9137 }, { "epoch": 0.9, "learning_rate": 5.11477319149598e-05, "loss": 1.3564, "step": 9138 }, { "epoch": 0.9, "learning_rate": 5.104697076216558e-05, "loss": 1.2783, "step": 9139 }, { "epoch": 0.9, "learning_rate": 5.094630635749187e-05, "loss": 1.3574, "step": 9140 }, { "epoch": 0.9, "learning_rate": 5.084573871120146e-05, "loss": 1.3623, "step": 9141 }, { "epoch": 0.9, "learning_rate": 5.074526783354794e-05, "loss": 1.3799, "step": 9142 }, { "epoch": 0.9, "learning_rate": 5.064489373477432e-05, "loss": 1.4453, "step": 9143 }, { "epoch": 0.9, "learning_rate": 5.0544616425113986e-05, "loss": 1.2598, "step": 9144 }, { "epoch": 0.9, "learning_rate": 5.044443591479053e-05, "loss": 1.4072, "step": 9145 }, { "epoch": 0.9, "learning_rate": 5.034435221401745e-05, "loss": 1.4258, "step": 9146 }, { "epoch": 0.9, "learning_rate": 5.02443653329987e-05, "loss": 1.5605, "step": 9147 }, { "epoch": 0.9, "learning_rate": 5.0144475281928004e-05, "loss": 1.374, "step": 9148 }, { "epoch": 0.9, "learning_rate": 5.0044682070989665e-05, "loss": 1.5137, "step": 9149 }, { "epoch": 0.9, "learning_rate": 4.994498571035755e-05, "loss": 1.3184, "step": 9150 }, { "epoch": 0.9, "learning_rate": 4.984538621019618e-05, "loss": 1.458, "step": 9151 }, { "epoch": 0.9, "learning_rate": 4.974588358065979e-05, "loss": 1.459, "step": 9152 }, { "epoch": 0.9, "learning_rate": 4.964647783189313e-05, "loss": 1.499, "step": 9153 }, { "epoch": 0.9, "learning_rate": 4.954716897403067e-05, "loss": 1.2959, "step": 9154 }, { "epoch": 0.9, "learning_rate": 4.944795701719729e-05, "loss": 1.4219, "step": 9155 }, { "epoch": 0.9, "learning_rate": 4.9348841971507905e-05, "loss": 1.4795, "step": 9156 }, { "epoch": 0.9, "learning_rate": 4.9249823847067554e-05, "loss": 1.5146, "step": 9157 }, { "epoch": 0.9, "learning_rate": 4.9150902653971265e-05, "loss": 1.2773, "step": 9158 }, { "epoch": 0.9, "learning_rate": 4.90520784023043e-05, "loss": 1.3887, "step": 9159 }, { "epoch": 0.9, "learning_rate": 4.895335110214216e-05, "loss": 1.3662, "step": 9160 }, { "epoch": 0.9, "learning_rate": 4.8854720763550464e-05, "loss": 1.415, "step": 9161 }, { "epoch": 0.9, "learning_rate": 4.8756187396584604e-05, "loss": 1.1963, "step": 9162 }, { "epoch": 0.9, "learning_rate": 4.865775101129022e-05, "loss": 1.269, "step": 9163 }, { "epoch": 0.9, "learning_rate": 4.855941161770339e-05, "loss": 1.3462, "step": 9164 }, { "epoch": 0.9, "learning_rate": 4.846116922584987e-05, "loss": 1.3174, "step": 9165 }, { "epoch": 0.9, "learning_rate": 4.8363023845745756e-05, "loss": 1.5371, "step": 9166 }, { "epoch": 0.9, "learning_rate": 4.826497548739739e-05, "loss": 1.3721, "step": 9167 }, { "epoch": 0.9, "learning_rate": 4.816702416080076e-05, "loss": 1.2646, "step": 9168 }, { "epoch": 0.9, "learning_rate": 4.806916987594245e-05, "loss": 1.4023, "step": 9169 }, { "epoch": 0.9, "learning_rate": 4.797141264279881e-05, "loss": 1.459, "step": 9170 }, { "epoch": 0.9, "learning_rate": 4.787375247133663e-05, "loss": 1.3789, "step": 9171 }, { "epoch": 0.9, "learning_rate": 4.77761893715124e-05, "loss": 1.3047, "step": 9172 }, { "epoch": 0.9, "learning_rate": 4.7678723353272944e-05, "loss": 1.4922, "step": 9173 }, { "epoch": 0.9, "learning_rate": 4.7581354426555315e-05, "loss": 1.4307, "step": 9174 }, { "epoch": 0.9, "learning_rate": 4.748408260128645e-05, "loss": 1.5195, "step": 9175 }, { "epoch": 0.9, "learning_rate": 4.7386907887383426e-05, "loss": 1.3643, "step": 9176 }, { "epoch": 0.9, "learning_rate": 4.728983029475342e-05, "loss": 1.3398, "step": 9177 }, { "epoch": 0.9, "learning_rate": 4.719284983329375e-05, "loss": 1.3066, "step": 9178 }, { "epoch": 0.9, "learning_rate": 4.709596651289194e-05, "loss": 1.4072, "step": 9179 }, { "epoch": 0.9, "learning_rate": 4.699918034342532e-05, "loss": 1.4258, "step": 9180 }, { "epoch": 0.91, "learning_rate": 4.6902491334761545e-05, "loss": 1.4658, "step": 9181 }, { "epoch": 0.91, "learning_rate": 4.6805899496758283e-05, "loss": 1.2427, "step": 9182 }, { "epoch": 0.91, "learning_rate": 4.670940483926334e-05, "loss": 1.2983, "step": 9183 }, { "epoch": 0.91, "learning_rate": 4.661300737211449e-05, "loss": 1.5059, "step": 9184 }, { "epoch": 0.91, "learning_rate": 4.651670710513989e-05, "loss": 1.4229, "step": 9185 }, { "epoch": 0.91, "learning_rate": 4.642050404815745e-05, "loss": 1.4805, "step": 9186 }, { "epoch": 0.91, "learning_rate": 4.6324398210975335e-05, "loss": 1.3428, "step": 9187 }, { "epoch": 0.91, "learning_rate": 4.622838960339171e-05, "loss": 1.3545, "step": 9188 }, { "epoch": 0.91, "learning_rate": 4.613247823519506e-05, "loss": 1.4385, "step": 9189 }, { "epoch": 0.91, "learning_rate": 4.603666411616358e-05, "loss": 1.2852, "step": 9190 }, { "epoch": 0.91, "learning_rate": 4.594094725606579e-05, "loss": 1.2949, "step": 9191 }, { "epoch": 0.91, "learning_rate": 4.584532766466043e-05, "loss": 1.3271, "step": 9192 }, { "epoch": 0.91, "learning_rate": 4.574980535169615e-05, "loss": 1.4756, "step": 9193 }, { "epoch": 0.91, "learning_rate": 4.565438032691149e-05, "loss": 1.3311, "step": 9194 }, { "epoch": 0.91, "learning_rate": 4.5559052600035454e-05, "loss": 1.4297, "step": 9195 }, { "epoch": 0.91, "learning_rate": 4.546382218078682e-05, "loss": 1.4717, "step": 9196 }, { "epoch": 0.91, "learning_rate": 4.536868907887459e-05, "loss": 1.4756, "step": 9197 }, { "epoch": 0.91, "learning_rate": 4.52736533039978e-05, "loss": 1.4404, "step": 9198 }, { "epoch": 0.91, "learning_rate": 4.5178714865845684e-05, "loss": 1.3965, "step": 9199 }, { "epoch": 0.91, "learning_rate": 4.5083873774097396e-05, "loss": 1.4844, "step": 9200 }, { "epoch": 0.91, "learning_rate": 4.49891300384222e-05, "loss": 1.374, "step": 9201 }, { "epoch": 0.91, "learning_rate": 4.489448366847948e-05, "loss": 1.3672, "step": 9202 }, { "epoch": 0.91, "learning_rate": 4.479993467391874e-05, "loss": 1.3809, "step": 9203 }, { "epoch": 0.91, "learning_rate": 4.470548306437927e-05, "loss": 1.376, "step": 9204 }, { "epoch": 0.91, "learning_rate": 4.4611128849490814e-05, "loss": 1.3984, "step": 9205 }, { "epoch": 0.91, "learning_rate": 4.4516872038873e-05, "loss": 1.3022, "step": 9206 }, { "epoch": 0.91, "learning_rate": 4.4422712642135484e-05, "loss": 1.3403, "step": 9207 }, { "epoch": 0.91, "learning_rate": 4.4328650668878145e-05, "loss": 1.4023, "step": 9208 }, { "epoch": 0.91, "learning_rate": 4.423468612869064e-05, "loss": 1.2798, "step": 9209 }, { "epoch": 0.91, "learning_rate": 4.414081903115319e-05, "loss": 1.4004, "step": 9210 }, { "epoch": 0.91, "learning_rate": 4.404704938583559e-05, "loss": 1.3672, "step": 9211 }, { "epoch": 0.91, "learning_rate": 4.395337720229797e-05, "loss": 1.418, "step": 9212 }, { "epoch": 0.91, "learning_rate": 4.385980249009036e-05, "loss": 1.4004, "step": 9213 }, { "epoch": 0.91, "learning_rate": 4.37663252587529e-05, "loss": 1.3926, "step": 9214 }, { "epoch": 0.91, "learning_rate": 4.367294551781598e-05, "loss": 1.2983, "step": 9215 }, { "epoch": 0.91, "learning_rate": 4.3579663276799855e-05, "loss": 1.4629, "step": 9216 }, { "epoch": 0.91, "learning_rate": 4.348647854521481e-05, "loss": 1.2373, "step": 9217 }, { "epoch": 0.91, "learning_rate": 4.339339133256126e-05, "loss": 1.499, "step": 9218 }, { "epoch": 0.91, "learning_rate": 4.330040164832971e-05, "loss": 1.4648, "step": 9219 }, { "epoch": 0.91, "learning_rate": 4.32075095020007e-05, "loss": 1.4688, "step": 9220 }, { "epoch": 0.91, "learning_rate": 4.311471490304475e-05, "loss": 1.2896, "step": 9221 }, { "epoch": 0.91, "learning_rate": 4.302201786092252e-05, "loss": 1.374, "step": 9222 }, { "epoch": 0.91, "learning_rate": 4.292941838508468e-05, "loss": 1.3555, "step": 9223 }, { "epoch": 0.91, "learning_rate": 4.283691648497212e-05, "loss": 1.3682, "step": 9224 }, { "epoch": 0.91, "learning_rate": 4.2744512170015515e-05, "loss": 1.5029, "step": 9225 }, { "epoch": 0.91, "learning_rate": 4.265220544963577e-05, "loss": 1.4619, "step": 9226 }, { "epoch": 0.91, "learning_rate": 4.255999633324381e-05, "loss": 1.3252, "step": 9227 }, { "epoch": 0.91, "learning_rate": 4.246788483024044e-05, "loss": 1.4609, "step": 9228 }, { "epoch": 0.91, "learning_rate": 4.2375870950016826e-05, "loss": 1.3447, "step": 9229 }, { "epoch": 0.91, "learning_rate": 4.2283954701953894e-05, "loss": 1.3936, "step": 9230 }, { "epoch": 0.91, "learning_rate": 4.2192136095422826e-05, "loss": 1.3418, "step": 9231 }, { "epoch": 0.91, "learning_rate": 4.2100415139784684e-05, "loss": 1.4072, "step": 9232 }, { "epoch": 0.91, "learning_rate": 4.200879184439077e-05, "loss": 1.4727, "step": 9233 }, { "epoch": 0.91, "learning_rate": 4.191726621858205e-05, "loss": 1.3613, "step": 9234 }, { "epoch": 0.91, "learning_rate": 4.1825838271690066e-05, "loss": 1.3447, "step": 9235 }, { "epoch": 0.91, "learning_rate": 4.173450801303591e-05, "loss": 1.3252, "step": 9236 }, { "epoch": 0.91, "learning_rate": 4.164327545193114e-05, "loss": 1.3496, "step": 9237 }, { "epoch": 0.91, "learning_rate": 4.155214059767698e-05, "loss": 1.4326, "step": 9238 }, { "epoch": 0.91, "learning_rate": 4.1461103459564885e-05, "loss": 1.3799, "step": 9239 }, { "epoch": 0.91, "learning_rate": 4.137016404687632e-05, "loss": 1.3809, "step": 9240 }, { "epoch": 0.91, "learning_rate": 4.1279322368882655e-05, "loss": 1.4375, "step": 9241 }, { "epoch": 0.91, "learning_rate": 4.1188578434845804e-05, "loss": 1.4639, "step": 9242 }, { "epoch": 0.91, "learning_rate": 4.1097932254017035e-05, "loss": 1.415, "step": 9243 }, { "epoch": 0.91, "learning_rate": 4.100738383563796e-05, "loss": 1.4053, "step": 9244 }, { "epoch": 0.91, "learning_rate": 4.091693318894041e-05, "loss": 1.4556, "step": 9245 }, { "epoch": 0.91, "learning_rate": 4.082658032314579e-05, "loss": 1.4326, "step": 9246 }, { "epoch": 0.91, "learning_rate": 4.073632524746607e-05, "loss": 1.3887, "step": 9247 }, { "epoch": 0.91, "learning_rate": 4.0646167971102654e-05, "loss": 1.3721, "step": 9248 }, { "epoch": 0.91, "learning_rate": 4.055610850324765e-05, "loss": 1.3398, "step": 9249 }, { "epoch": 0.91, "learning_rate": 4.0466146853082587e-05, "loss": 1.415, "step": 9250 }, { "epoch": 0.91, "learning_rate": 4.0376283029779356e-05, "loss": 1.4561, "step": 9251 }, { "epoch": 0.91, "learning_rate": 4.028651704249986e-05, "loss": 1.4482, "step": 9252 }, { "epoch": 0.91, "learning_rate": 4.019684890039588e-05, "loss": 1.4326, "step": 9253 }, { "epoch": 0.91, "learning_rate": 4.010727861260943e-05, "loss": 1.4883, "step": 9254 }, { "epoch": 0.91, "learning_rate": 4.001780618827222e-05, "loss": 1.4355, "step": 9255 }, { "epoch": 0.91, "learning_rate": 3.992843163650639e-05, "loss": 1.4424, "step": 9256 }, { "epoch": 0.91, "learning_rate": 3.983915496642376e-05, "loss": 1.5059, "step": 9257 }, { "epoch": 0.91, "learning_rate": 3.974997618712639e-05, "loss": 1.3672, "step": 9258 }, { "epoch": 0.91, "learning_rate": 3.9660895307706116e-05, "loss": 1.2651, "step": 9259 }, { "epoch": 0.91, "learning_rate": 3.957191233724522e-05, "loss": 1.3711, "step": 9260 }, { "epoch": 0.91, "learning_rate": 3.948302728481556e-05, "loss": 1.2812, "step": 9261 }, { "epoch": 0.91, "learning_rate": 3.9394240159479234e-05, "loss": 1.2783, "step": 9262 }, { "epoch": 0.91, "learning_rate": 3.930555097028832e-05, "loss": 1.3271, "step": 9263 }, { "epoch": 0.91, "learning_rate": 3.921695972628492e-05, "loss": 1.3252, "step": 9264 }, { "epoch": 0.91, "learning_rate": 3.912846643650103e-05, "loss": 1.4736, "step": 9265 }, { "epoch": 0.91, "learning_rate": 3.904007110995888e-05, "loss": 1.376, "step": 9266 }, { "epoch": 0.91, "learning_rate": 3.895177375567049e-05, "loss": 1.4209, "step": 9267 }, { "epoch": 0.91, "learning_rate": 3.88635743826381e-05, "loss": 1.3105, "step": 9268 }, { "epoch": 0.91, "learning_rate": 3.8775472999853844e-05, "loss": 1.3613, "step": 9269 }, { "epoch": 0.91, "learning_rate": 3.868746961629976e-05, "loss": 1.5596, "step": 9270 }, { "epoch": 0.91, "learning_rate": 3.859956424094813e-05, "loss": 1.3525, "step": 9271 }, { "epoch": 0.91, "learning_rate": 3.8511756882761094e-05, "loss": 1.4492, "step": 9272 }, { "epoch": 0.91, "learning_rate": 3.8424047550690846e-05, "loss": 1.4062, "step": 9273 }, { "epoch": 0.91, "learning_rate": 3.833643625367944e-05, "loss": 1.3652, "step": 9274 }, { "epoch": 0.91, "learning_rate": 3.824892300065918e-05, "loss": 1.3838, "step": 9275 }, { "epoch": 0.91, "learning_rate": 3.8161507800552275e-05, "loss": 1.2832, "step": 9276 }, { "epoch": 0.91, "learning_rate": 3.807419066227091e-05, "loss": 1.3828, "step": 9277 }, { "epoch": 0.91, "learning_rate": 3.798697159471731e-05, "loss": 1.4658, "step": 9278 }, { "epoch": 0.91, "learning_rate": 3.789985060678358e-05, "loss": 1.3828, "step": 9279 }, { "epoch": 0.91, "learning_rate": 3.781282770735206e-05, "loss": 1.4355, "step": 9280 }, { "epoch": 0.91, "learning_rate": 3.772590290529476e-05, "loss": 1.4834, "step": 9281 }, { "epoch": 0.92, "learning_rate": 3.7639076209474044e-05, "loss": 1.3179, "step": 9282 }, { "epoch": 0.92, "learning_rate": 3.7552347628742045e-05, "loss": 1.3701, "step": 9283 }, { "epoch": 0.92, "learning_rate": 3.746571717194103e-05, "loss": 1.4482, "step": 9284 }, { "epoch": 0.92, "learning_rate": 3.737918484790304e-05, "loss": 1.4014, "step": 9285 }, { "epoch": 0.92, "learning_rate": 3.729275066545035e-05, "loss": 1.502, "step": 9286 }, { "epoch": 0.92, "learning_rate": 3.720641463339513e-05, "loss": 1.3555, "step": 9287 }, { "epoch": 0.92, "learning_rate": 3.7120176760539694e-05, "loss": 1.3457, "step": 9288 }, { "epoch": 0.92, "learning_rate": 3.703403705567598e-05, "loss": 1.3135, "step": 9289 }, { "epoch": 0.92, "learning_rate": 3.694799552758621e-05, "loss": 1.4385, "step": 9290 }, { "epoch": 0.92, "learning_rate": 3.686205218504257e-05, "loss": 1.3979, "step": 9291 }, { "epoch": 0.92, "learning_rate": 3.6776207036807066e-05, "loss": 1.3906, "step": 9292 }, { "epoch": 0.92, "learning_rate": 3.669046009163213e-05, "loss": 1.4463, "step": 9293 }, { "epoch": 0.92, "learning_rate": 3.660481135825966e-05, "loss": 1.5596, "step": 9294 }, { "epoch": 0.92, "learning_rate": 3.651926084542168e-05, "loss": 1.3965, "step": 9295 }, { "epoch": 0.92, "learning_rate": 3.643380856184053e-05, "loss": 1.3481, "step": 9296 }, { "epoch": 0.92, "learning_rate": 3.634845451622804e-05, "loss": 1.2925, "step": 9297 }, { "epoch": 0.92, "learning_rate": 3.626319871728645e-05, "loss": 1.3984, "step": 9298 }, { "epoch": 0.92, "learning_rate": 3.617804117370771e-05, "loss": 1.3896, "step": 9299 }, { "epoch": 0.92, "learning_rate": 3.609298189417376e-05, "loss": 1.4004, "step": 9300 }, { "epoch": 0.92, "learning_rate": 3.600802088735678e-05, "loss": 1.3623, "step": 9301 }, { "epoch": 0.92, "learning_rate": 3.592315816191871e-05, "loss": 1.4287, "step": 9302 }, { "epoch": 0.92, "learning_rate": 3.5838393726511434e-05, "loss": 1.271, "step": 9303 }, { "epoch": 0.92, "learning_rate": 3.575372758977691e-05, "loss": 1.5264, "step": 9304 }, { "epoch": 0.92, "learning_rate": 3.566915976034712e-05, "loss": 1.2637, "step": 9305 }, { "epoch": 0.92, "learning_rate": 3.558469024684386e-05, "loss": 1.4014, "step": 9306 }, { "epoch": 0.92, "learning_rate": 3.5500319057879206e-05, "loss": 1.4346, "step": 9307 }, { "epoch": 0.92, "learning_rate": 3.541604620205485e-05, "loss": 1.4326, "step": 9308 }, { "epoch": 0.92, "learning_rate": 3.533187168796259e-05, "loss": 1.3374, "step": 9309 }, { "epoch": 0.92, "learning_rate": 3.524779552418422e-05, "loss": 1.4795, "step": 9310 }, { "epoch": 0.92, "learning_rate": 3.516381771929167e-05, "loss": 1.4111, "step": 9311 }, { "epoch": 0.92, "learning_rate": 3.5079938281846744e-05, "loss": 1.3721, "step": 9312 }, { "epoch": 0.92, "learning_rate": 3.499615722040084e-05, "loss": 1.3242, "step": 9313 }, { "epoch": 0.92, "learning_rate": 3.491247454349589e-05, "loss": 1.2979, "step": 9314 }, { "epoch": 0.92, "learning_rate": 3.4828890259663514e-05, "loss": 1.4072, "step": 9315 }, { "epoch": 0.92, "learning_rate": 3.474540437742535e-05, "loss": 1.5059, "step": 9316 }, { "epoch": 0.92, "learning_rate": 3.466201690529281e-05, "loss": 1.3467, "step": 9317 }, { "epoch": 0.92, "learning_rate": 3.457872785176774e-05, "loss": 1.3105, "step": 9318 }, { "epoch": 0.92, "learning_rate": 3.4495537225341376e-05, "loss": 1.3457, "step": 9319 }, { "epoch": 0.92, "learning_rate": 3.441244503449548e-05, "loss": 1.4307, "step": 9320 }, { "epoch": 0.92, "learning_rate": 3.432945128770137e-05, "loss": 1.3828, "step": 9321 }, { "epoch": 0.92, "learning_rate": 3.4246555993420415e-05, "loss": 1.457, "step": 9322 }, { "epoch": 0.92, "learning_rate": 3.4163759160104054e-05, "loss": 1.335, "step": 9323 }, { "epoch": 0.92, "learning_rate": 3.408106079619366e-05, "loss": 1.5322, "step": 9324 }, { "epoch": 0.92, "learning_rate": 3.399846091012049e-05, "loss": 1.4795, "step": 9325 }, { "epoch": 0.92, "learning_rate": 3.39159595103059e-05, "loss": 1.2588, "step": 9326 }, { "epoch": 0.92, "learning_rate": 3.383355660516096e-05, "loss": 1.3701, "step": 9327 }, { "epoch": 0.92, "learning_rate": 3.3751252203086834e-05, "loss": 1.4072, "step": 9328 }, { "epoch": 0.92, "learning_rate": 3.3669046312475026e-05, "loss": 1.5312, "step": 9329 }, { "epoch": 0.92, "learning_rate": 3.358693894170617e-05, "loss": 1.4224, "step": 9330 }, { "epoch": 0.92, "learning_rate": 3.350493009915156e-05, "loss": 1.3584, "step": 9331 }, { "epoch": 0.92, "learning_rate": 3.3423019793172196e-05, "loss": 1.3896, "step": 9332 }, { "epoch": 0.92, "learning_rate": 3.334120803211893e-05, "loss": 1.5029, "step": 9333 }, { "epoch": 0.92, "learning_rate": 3.325949482433277e-05, "loss": 1.4189, "step": 9334 }, { "epoch": 0.92, "learning_rate": 3.317788017814449e-05, "loss": 1.3496, "step": 9335 }, { "epoch": 0.92, "learning_rate": 3.3096364101874996e-05, "loss": 1.4307, "step": 9336 }, { "epoch": 0.92, "learning_rate": 3.3014946603835084e-05, "loss": 1.4355, "step": 9337 }, { "epoch": 0.92, "learning_rate": 3.293362769232533e-05, "loss": 1.4912, "step": 9338 }, { "epoch": 0.92, "learning_rate": 3.285240737563644e-05, "loss": 1.3516, "step": 9339 }, { "epoch": 0.92, "learning_rate": 3.277128566204912e-05, "loss": 1.5381, "step": 9340 }, { "epoch": 0.92, "learning_rate": 3.2690262559833874e-05, "loss": 1.3877, "step": 9341 }, { "epoch": 0.92, "learning_rate": 3.260933807725119e-05, "loss": 1.3594, "step": 9342 }, { "epoch": 0.92, "learning_rate": 3.252851222255149e-05, "loss": 1.4746, "step": 9343 }, { "epoch": 0.92, "learning_rate": 3.244778500397527e-05, "loss": 1.4219, "step": 9344 }, { "epoch": 0.92, "learning_rate": 3.2367156429752856e-05, "loss": 1.375, "step": 9345 }, { "epoch": 0.92, "learning_rate": 3.2286626508104546e-05, "loss": 1.2837, "step": 9346 }, { "epoch": 0.92, "learning_rate": 3.220619524724044e-05, "loss": 1.2725, "step": 9347 }, { "epoch": 0.92, "learning_rate": 3.212586265536077e-05, "loss": 1.54, "step": 9348 }, { "epoch": 0.92, "learning_rate": 3.204562874065575e-05, "loss": 1.2637, "step": 9349 }, { "epoch": 0.92, "learning_rate": 3.1965493511305176e-05, "loss": 1.2998, "step": 9350 }, { "epoch": 0.92, "learning_rate": 3.1885456975479284e-05, "loss": 1.3809, "step": 9351 }, { "epoch": 0.92, "learning_rate": 3.1805519141337894e-05, "loss": 1.4131, "step": 9352 }, { "epoch": 0.92, "learning_rate": 3.172568001703091e-05, "loss": 1.2441, "step": 9353 }, { "epoch": 0.92, "learning_rate": 3.164593961069806e-05, "loss": 1.4756, "step": 9354 }, { "epoch": 0.92, "learning_rate": 3.156629793046917e-05, "loss": 1.3975, "step": 9355 }, { "epoch": 0.92, "learning_rate": 3.148675498446374e-05, "loss": 1.3652, "step": 9356 }, { "epoch": 0.92, "learning_rate": 3.140731078079162e-05, "loss": 1.3145, "step": 9357 }, { "epoch": 0.92, "learning_rate": 3.132796532755211e-05, "loss": 1.3057, "step": 9358 }, { "epoch": 0.92, "learning_rate": 3.124871863283474e-05, "loss": 1.4375, "step": 9359 }, { "epoch": 0.92, "learning_rate": 3.116957070471893e-05, "loss": 1.4141, "step": 9360 }, { "epoch": 0.92, "learning_rate": 3.1090521551274007e-05, "loss": 1.4688, "step": 9361 }, { "epoch": 0.92, "learning_rate": 3.10115711805593e-05, "loss": 1.4756, "step": 9362 }, { "epoch": 0.92, "learning_rate": 3.0932719600623936e-05, "loss": 1.3462, "step": 9363 }, { "epoch": 0.92, "learning_rate": 3.0853966819507026e-05, "loss": 1.3682, "step": 9364 }, { "epoch": 0.92, "learning_rate": 3.07753128452376e-05, "loss": 1.333, "step": 9365 }, { "epoch": 0.92, "learning_rate": 3.0696757685834686e-05, "loss": 1.4399, "step": 9366 }, { "epoch": 0.92, "learning_rate": 3.06183013493071e-05, "loss": 1.4883, "step": 9367 }, { "epoch": 0.92, "learning_rate": 3.0539943843653773e-05, "loss": 1.4287, "step": 9368 }, { "epoch": 0.92, "learning_rate": 3.0461685176863206e-05, "loss": 1.4551, "step": 9369 }, { "epoch": 0.92, "learning_rate": 3.0383525356914354e-05, "loss": 1.3408, "step": 9370 }, { "epoch": 0.92, "learning_rate": 3.0305464391775616e-05, "loss": 1.46, "step": 9371 }, { "epoch": 0.92, "learning_rate": 3.0227502289405517e-05, "loss": 1.3232, "step": 9372 }, { "epoch": 0.92, "learning_rate": 3.01496390577527e-05, "loss": 1.2451, "step": 9373 }, { "epoch": 0.92, "learning_rate": 3.007187470475514e-05, "loss": 1.4219, "step": 9374 }, { "epoch": 0.92, "learning_rate": 2.9994209238341508e-05, "loss": 1.3936, "step": 9375 }, { "epoch": 0.92, "learning_rate": 2.9916642666429682e-05, "loss": 1.4756, "step": 9376 }, { "epoch": 0.92, "learning_rate": 2.98391749969279e-05, "loss": 1.5576, "step": 9377 }, { "epoch": 0.92, "learning_rate": 2.9761806237734057e-05, "loss": 1.3369, "step": 9378 }, { "epoch": 0.92, "learning_rate": 2.9684536396736296e-05, "loss": 1.4355, "step": 9379 }, { "epoch": 0.92, "learning_rate": 2.9607365481812308e-05, "loss": 1.1855, "step": 9380 }, { "epoch": 0.92, "learning_rate": 2.953029350083003e-05, "loss": 1.3994, "step": 9381 }, { "epoch": 0.92, "learning_rate": 2.945332046164706e-05, "loss": 1.3281, "step": 9382 }, { "epoch": 0.92, "learning_rate": 2.93764463721109e-05, "loss": 1.3799, "step": 9383 }, { "epoch": 0.93, "learning_rate": 2.9299671240059055e-05, "loss": 1.4443, "step": 9384 }, { "epoch": 0.93, "learning_rate": 2.922299507331916e-05, "loss": 1.5273, "step": 9385 }, { "epoch": 0.93, "learning_rate": 2.9146417879708398e-05, "loss": 1.3564, "step": 9386 }, { "epoch": 0.93, "learning_rate": 2.9069939667033863e-05, "loss": 1.4385, "step": 9387 }, { "epoch": 0.93, "learning_rate": 2.899356044309287e-05, "loss": 1.376, "step": 9388 }, { "epoch": 0.93, "learning_rate": 2.8917280215672414e-05, "loss": 1.4961, "step": 9389 }, { "epoch": 0.93, "learning_rate": 2.8841098992549496e-05, "loss": 1.3838, "step": 9390 }, { "epoch": 0.93, "learning_rate": 2.8765016781490904e-05, "loss": 1.5137, "step": 9391 }, { "epoch": 0.93, "learning_rate": 2.868903359025343e-05, "loss": 1.3291, "step": 9392 }, { "epoch": 0.93, "learning_rate": 2.8613149426583884e-05, "loss": 1.3848, "step": 9393 }, { "epoch": 0.93, "learning_rate": 2.8537364298218627e-05, "loss": 1.2651, "step": 9394 }, { "epoch": 0.93, "learning_rate": 2.846167821288437e-05, "loss": 1.4209, "step": 9395 }, { "epoch": 0.93, "learning_rate": 2.838609117829727e-05, "loss": 1.2832, "step": 9396 }, { "epoch": 0.93, "learning_rate": 2.831060320216383e-05, "loss": 1.3662, "step": 9397 }, { "epoch": 0.93, "learning_rate": 2.8235214292180124e-05, "loss": 1.3936, "step": 9398 }, { "epoch": 0.93, "learning_rate": 2.8159924456032215e-05, "loss": 1.4395, "step": 9399 }, { "epoch": 0.93, "learning_rate": 2.808473370139608e-05, "loss": 1.3408, "step": 9400 }, { "epoch": 0.93, "learning_rate": 2.80096420359377e-05, "loss": 1.376, "step": 9401 }, { "epoch": 0.93, "learning_rate": 2.7934649467312833e-05, "loss": 1.4951, "step": 9402 }, { "epoch": 0.93, "learning_rate": 2.7859756003167036e-05, "loss": 1.3179, "step": 9403 }, { "epoch": 0.93, "learning_rate": 2.7784961651135977e-05, "loss": 1.3438, "step": 9404 }, { "epoch": 0.93, "learning_rate": 2.771026641884522e-05, "loss": 1.3042, "step": 9405 }, { "epoch": 0.93, "learning_rate": 2.7635670313910012e-05, "loss": 1.3447, "step": 9406 }, { "epoch": 0.93, "learning_rate": 2.7561173343935596e-05, "loss": 1.335, "step": 9407 }, { "epoch": 0.93, "learning_rate": 2.7486775516517128e-05, "loss": 1.4351, "step": 9408 }, { "epoch": 0.93, "learning_rate": 2.741247683923964e-05, "loss": 1.3496, "step": 9409 }, { "epoch": 0.93, "learning_rate": 2.7338277319678084e-05, "loss": 1.3672, "step": 9410 }, { "epoch": 0.93, "learning_rate": 2.7264176965397402e-05, "loss": 1.3525, "step": 9411 }, { "epoch": 0.93, "learning_rate": 2.719017578395222e-05, "loss": 1.4004, "step": 9412 }, { "epoch": 0.93, "learning_rate": 2.711627378288717e-05, "loss": 1.4023, "step": 9413 }, { "epoch": 0.93, "learning_rate": 2.7042470969736666e-05, "loss": 1.3457, "step": 9414 }, { "epoch": 0.93, "learning_rate": 2.6968767352025248e-05, "loss": 1.3774, "step": 9415 }, { "epoch": 0.93, "learning_rate": 2.6895162937266905e-05, "loss": 1.3779, "step": 9416 }, { "epoch": 0.93, "learning_rate": 2.6821657732966076e-05, "loss": 1.3887, "step": 9417 }, { "epoch": 0.93, "learning_rate": 2.6748251746616658e-05, "loss": 1.2754, "step": 9418 }, { "epoch": 0.93, "learning_rate": 2.6674944985702552e-05, "loss": 1.3018, "step": 9419 }, { "epoch": 0.93, "learning_rate": 2.6601737457697673e-05, "loss": 1.4297, "step": 9420 }, { "epoch": 0.93, "learning_rate": 2.65286291700656e-05, "loss": 1.3496, "step": 9421 }, { "epoch": 0.93, "learning_rate": 2.6455620130259928e-05, "loss": 1.4141, "step": 9422 }, { "epoch": 0.93, "learning_rate": 2.638271034572415e-05, "loss": 1.2896, "step": 9423 }, { "epoch": 0.93, "learning_rate": 2.630989982389165e-05, "loss": 1.4414, "step": 9424 }, { "epoch": 0.93, "learning_rate": 2.6237188572185488e-05, "loss": 1.27, "step": 9425 }, { "epoch": 0.93, "learning_rate": 2.6164576598018852e-05, "loss": 1.457, "step": 9426 }, { "epoch": 0.93, "learning_rate": 2.6092063908794818e-05, "loss": 1.3804, "step": 9427 }, { "epoch": 0.93, "learning_rate": 2.601965051190591e-05, "loss": 1.3359, "step": 9428 }, { "epoch": 0.93, "learning_rate": 2.5947336414735124e-05, "loss": 1.3999, "step": 9429 }, { "epoch": 0.93, "learning_rate": 2.5875121624655106e-05, "loss": 1.5039, "step": 9430 }, { "epoch": 0.93, "learning_rate": 2.5803006149028196e-05, "loss": 1.3496, "step": 9431 }, { "epoch": 0.93, "learning_rate": 2.5730989995206845e-05, "loss": 1.2666, "step": 9432 }, { "epoch": 0.93, "learning_rate": 2.5659073170533064e-05, "loss": 1.4287, "step": 9433 }, { "epoch": 0.93, "learning_rate": 2.5587255682339217e-05, "loss": 1.5068, "step": 9434 }, { "epoch": 0.93, "learning_rate": 2.551553753794711e-05, "loss": 1.3027, "step": 9435 }, { "epoch": 0.93, "learning_rate": 2.544391874466867e-05, "loss": 1.3379, "step": 9436 }, { "epoch": 0.93, "learning_rate": 2.5372399309805616e-05, "loss": 1.4092, "step": 9437 }, { "epoch": 0.93, "learning_rate": 2.5300979240649446e-05, "loss": 1.3267, "step": 9438 }, { "epoch": 0.93, "learning_rate": 2.522965854448167e-05, "loss": 1.375, "step": 9439 }, { "epoch": 0.93, "learning_rate": 2.5158437228573583e-05, "loss": 1.2705, "step": 9440 }, { "epoch": 0.93, "learning_rate": 2.5087315300186485e-05, "loss": 1.4062, "step": 9441 }, { "epoch": 0.93, "learning_rate": 2.5016292766571246e-05, "loss": 1.3779, "step": 9442 }, { "epoch": 0.93, "learning_rate": 2.4945369634968852e-05, "loss": 1.3613, "step": 9443 }, { "epoch": 0.93, "learning_rate": 2.487454591261018e-05, "loss": 1.375, "step": 9444 }, { "epoch": 0.93, "learning_rate": 2.4803821606715902e-05, "loss": 1.4268, "step": 9445 }, { "epoch": 0.93, "learning_rate": 2.4733196724496365e-05, "loss": 1.3091, "step": 9446 }, { "epoch": 0.93, "learning_rate": 2.4662671273152138e-05, "loss": 1.3857, "step": 9447 }, { "epoch": 0.93, "learning_rate": 2.459224525987336e-05, "loss": 1.3223, "step": 9448 }, { "epoch": 0.93, "learning_rate": 2.452191869184006e-05, "loss": 1.2979, "step": 9449 }, { "epoch": 0.93, "learning_rate": 2.4451691576222403e-05, "loss": 1.4434, "step": 9450 }, { "epoch": 0.93, "learning_rate": 2.4381563920180093e-05, "loss": 1.4053, "step": 9451 }, { "epoch": 0.93, "learning_rate": 2.4311535730862865e-05, "loss": 1.4395, "step": 9452 }, { "epoch": 0.93, "learning_rate": 2.424160701541034e-05, "loss": 1.375, "step": 9453 }, { "epoch": 0.93, "learning_rate": 2.4171777780951697e-05, "loss": 1.3545, "step": 9454 }, { "epoch": 0.93, "learning_rate": 2.410204803460636e-05, "loss": 1.4014, "step": 9455 }, { "epoch": 0.93, "learning_rate": 2.4032417783483416e-05, "loss": 1.3745, "step": 9456 }, { "epoch": 0.93, "learning_rate": 2.396288703468197e-05, "loss": 1.3789, "step": 9457 }, { "epoch": 0.93, "learning_rate": 2.3893455795290565e-05, "loss": 1.3916, "step": 9458 }, { "epoch": 0.93, "learning_rate": 2.3824124072388208e-05, "loss": 1.3413, "step": 9459 }, { "epoch": 0.93, "learning_rate": 2.3754891873043138e-05, "loss": 1.2666, "step": 9460 }, { "epoch": 0.93, "learning_rate": 2.368575920431393e-05, "loss": 1.4141, "step": 9461 }, { "epoch": 0.93, "learning_rate": 2.361672607324894e-05, "loss": 1.4609, "step": 9462 }, { "epoch": 0.93, "learning_rate": 2.3547792486886098e-05, "loss": 1.335, "step": 9463 }, { "epoch": 0.93, "learning_rate": 2.3478958452253342e-05, "loss": 1.2764, "step": 9464 }, { "epoch": 0.93, "learning_rate": 2.3410223976368493e-05, "loss": 1.4619, "step": 9465 }, { "epoch": 0.93, "learning_rate": 2.3341589066239288e-05, "loss": 1.418, "step": 9466 }, { "epoch": 0.93, "learning_rate": 2.3273053728863124e-05, "loss": 1.3618, "step": 9467 }, { "epoch": 0.93, "learning_rate": 2.3204617971227525e-05, "loss": 1.208, "step": 9468 }, { "epoch": 0.93, "learning_rate": 2.3136281800309467e-05, "loss": 1.3916, "step": 9469 }, { "epoch": 0.93, "learning_rate": 2.306804522307604e-05, "loss": 1.377, "step": 9470 }, { "epoch": 0.93, "learning_rate": 2.2999908246484235e-05, "loss": 1.5371, "step": 9471 }, { "epoch": 0.93, "learning_rate": 2.293187087748072e-05, "loss": 1.4238, "step": 9472 }, { "epoch": 0.93, "learning_rate": 2.2863933123002055e-05, "loss": 1.542, "step": 9473 }, { "epoch": 0.93, "learning_rate": 2.2796094989974704e-05, "loss": 1.3018, "step": 9474 }, { "epoch": 0.93, "learning_rate": 2.2728356485314904e-05, "loss": 1.3481, "step": 9475 }, { "epoch": 0.93, "learning_rate": 2.2660717615928693e-05, "loss": 1.5195, "step": 9476 }, { "epoch": 0.93, "learning_rate": 2.259317838871222e-05, "loss": 1.3564, "step": 9477 }, { "epoch": 0.93, "learning_rate": 2.252573881055109e-05, "loss": 1.3477, "step": 9478 }, { "epoch": 0.93, "learning_rate": 2.245839888832102e-05, "loss": 1.29, "step": 9479 }, { "epoch": 0.93, "learning_rate": 2.239115862888752e-05, "loss": 1.2319, "step": 9480 }, { "epoch": 0.93, "learning_rate": 2.2324018039105774e-05, "loss": 1.3291, "step": 9481 }, { "epoch": 0.93, "learning_rate": 2.2256977125821087e-05, "loss": 1.2793, "step": 9482 }, { "epoch": 0.93, "learning_rate": 2.2190035895868322e-05, "loss": 1.2637, "step": 9483 }, { "epoch": 0.93, "learning_rate": 2.2123194356072463e-05, "loss": 1.4766, "step": 9484 }, { "epoch": 0.94, "learning_rate": 2.205645251324795e-05, "loss": 1.4688, "step": 9485 }, { "epoch": 0.94, "learning_rate": 2.198981037419945e-05, "loss": 1.4229, "step": 9486 }, { "epoch": 0.94, "learning_rate": 2.1923267945721193e-05, "loss": 1.3975, "step": 9487 }, { "epoch": 0.94, "learning_rate": 2.1856825234597422e-05, "loss": 1.3311, "step": 9488 }, { "epoch": 0.94, "learning_rate": 2.179048224760205e-05, "loss": 1.3994, "step": 9489 }, { "epoch": 0.94, "learning_rate": 2.1724238991499002e-05, "loss": 1.4111, "step": 9490 }, { "epoch": 0.94, "learning_rate": 2.165809547304187e-05, "loss": 1.3916, "step": 9491 }, { "epoch": 0.94, "learning_rate": 2.159205169897416e-05, "loss": 1.3984, "step": 9492 }, { "epoch": 0.94, "learning_rate": 2.152610767602925e-05, "loss": 1.374, "step": 9493 }, { "epoch": 0.94, "learning_rate": 2.1460263410930324e-05, "loss": 1.3799, "step": 9494 }, { "epoch": 0.94, "learning_rate": 2.139451891039024e-05, "loss": 1.3828, "step": 9495 }, { "epoch": 0.94, "learning_rate": 2.1328874181111847e-05, "loss": 1.3594, "step": 9496 }, { "epoch": 0.94, "learning_rate": 2.126332922978791e-05, "loss": 1.3989, "step": 9497 }, { "epoch": 0.94, "learning_rate": 2.119788406310086e-05, "loss": 1.3154, "step": 9498 }, { "epoch": 0.94, "learning_rate": 2.1132538687722915e-05, "loss": 1.4023, "step": 9499 }, { "epoch": 0.94, "learning_rate": 2.1067293110316187e-05, "loss": 1.4873, "step": 9500 }, { "epoch": 0.94, "learning_rate": 2.1002147337532694e-05, "loss": 1.3857, "step": 9501 }, { "epoch": 0.94, "learning_rate": 2.0937101376014123e-05, "loss": 1.2212, "step": 9502 }, { "epoch": 0.94, "learning_rate": 2.0872155232392275e-05, "loss": 1.3994, "step": 9503 }, { "epoch": 0.94, "learning_rate": 2.0807308913288303e-05, "loss": 1.3008, "step": 9504 }, { "epoch": 0.94, "learning_rate": 2.0742562425313582e-05, "loss": 1.3623, "step": 9505 }, { "epoch": 0.94, "learning_rate": 2.0677915775069277e-05, "loss": 1.4541, "step": 9506 }, { "epoch": 0.94, "learning_rate": 2.0613368969146117e-05, "loss": 1.4707, "step": 9507 }, { "epoch": 0.94, "learning_rate": 2.054892201412484e-05, "loss": 1.3848, "step": 9508 }, { "epoch": 0.94, "learning_rate": 2.048457491657607e-05, "loss": 1.3369, "step": 9509 }, { "epoch": 0.94, "learning_rate": 2.0420327683060015e-05, "loss": 1.3691, "step": 9510 }, { "epoch": 0.94, "learning_rate": 2.0356180320126872e-05, "loss": 1.6055, "step": 9511 }, { "epoch": 0.94, "learning_rate": 2.0292132834316634e-05, "loss": 1.4775, "step": 9512 }, { "epoch": 0.94, "learning_rate": 2.0228185232159303e-05, "loss": 1.3828, "step": 9513 }, { "epoch": 0.94, "learning_rate": 2.0164337520174213e-05, "loss": 1.54, "step": 9514 }, { "epoch": 0.94, "learning_rate": 2.0100589704870832e-05, "loss": 1.3281, "step": 9515 }, { "epoch": 0.94, "learning_rate": 2.0036941792748618e-05, "loss": 1.2188, "step": 9516 }, { "epoch": 0.94, "learning_rate": 1.9973393790296392e-05, "loss": 1.416, "step": 9517 }, { "epoch": 0.94, "learning_rate": 1.9909945703993183e-05, "loss": 1.3174, "step": 9518 }, { "epoch": 0.94, "learning_rate": 1.984659754030771e-05, "loss": 1.2705, "step": 9519 }, { "epoch": 0.94, "learning_rate": 1.978334930569825e-05, "loss": 1.543, "step": 9520 }, { "epoch": 0.94, "learning_rate": 1.9720201006613316e-05, "loss": 1.3721, "step": 9521 }, { "epoch": 0.94, "learning_rate": 1.9657152649491082e-05, "loss": 1.4414, "step": 9522 }, { "epoch": 0.94, "learning_rate": 1.9594204240759307e-05, "loss": 1.3213, "step": 9523 }, { "epoch": 0.94, "learning_rate": 1.9531355786835736e-05, "loss": 1.4092, "step": 9524 }, { "epoch": 0.94, "learning_rate": 1.9468607294128136e-05, "loss": 1.4912, "step": 9525 }, { "epoch": 0.94, "learning_rate": 1.9405958769033727e-05, "loss": 1.3428, "step": 9526 }, { "epoch": 0.94, "learning_rate": 1.9343410217939618e-05, "loss": 1.2002, "step": 9527 }, { "epoch": 0.94, "learning_rate": 1.9280961647222927e-05, "loss": 1.4053, "step": 9528 }, { "epoch": 0.94, "learning_rate": 1.9218613063250344e-05, "loss": 1.3623, "step": 9529 }, { "epoch": 0.94, "learning_rate": 1.915636447237845e-05, "loss": 1.3379, "step": 9530 }, { "epoch": 0.94, "learning_rate": 1.9094215880953837e-05, "loss": 1.3516, "step": 9531 }, { "epoch": 0.94, "learning_rate": 1.903216729531243e-05, "loss": 1.4346, "step": 9532 }, { "epoch": 0.94, "learning_rate": 1.8970218721780508e-05, "loss": 1.2827, "step": 9533 }, { "epoch": 0.94, "learning_rate": 1.8908370166673683e-05, "loss": 1.3589, "step": 9534 }, { "epoch": 0.94, "learning_rate": 1.8846621636297578e-05, "loss": 1.4336, "step": 9535 }, { "epoch": 0.94, "learning_rate": 1.8784973136947714e-05, "loss": 1.4043, "step": 9536 }, { "epoch": 0.94, "learning_rate": 1.8723424674909283e-05, "loss": 1.2017, "step": 9537 }, { "epoch": 0.94, "learning_rate": 1.866197625645727e-05, "loss": 1.3535, "step": 9538 }, { "epoch": 0.94, "learning_rate": 1.8600627887856547e-05, "loss": 1.3857, "step": 9539 }, { "epoch": 0.94, "learning_rate": 1.853937957536156e-05, "loss": 1.4136, "step": 9540 }, { "epoch": 0.94, "learning_rate": 1.8478231325216976e-05, "loss": 1.3184, "step": 9541 }, { "epoch": 0.94, "learning_rate": 1.8417183143656812e-05, "loss": 1.3164, "step": 9542 }, { "epoch": 0.94, "learning_rate": 1.83562350369052e-05, "loss": 1.314, "step": 9543 }, { "epoch": 0.94, "learning_rate": 1.8295387011175945e-05, "loss": 1.3994, "step": 9544 }, { "epoch": 0.94, "learning_rate": 1.8234639072672533e-05, "loss": 1.4277, "step": 9545 }, { "epoch": 0.94, "learning_rate": 1.817399122758856e-05, "loss": 1.3887, "step": 9546 }, { "epoch": 0.94, "learning_rate": 1.8113443482106972e-05, "loss": 1.3408, "step": 9547 }, { "epoch": 0.94, "learning_rate": 1.8052995842400942e-05, "loss": 1.5732, "step": 9548 }, { "epoch": 0.94, "learning_rate": 1.7992648314633318e-05, "loss": 1.3906, "step": 9549 }, { "epoch": 0.94, "learning_rate": 1.7932400904956626e-05, "loss": 1.3242, "step": 9550 }, { "epoch": 0.94, "learning_rate": 1.7872253619513057e-05, "loss": 1.3369, "step": 9551 }, { "epoch": 0.94, "learning_rate": 1.7812206464435045e-05, "loss": 1.3389, "step": 9552 }, { "epoch": 0.94, "learning_rate": 1.7752259445844354e-05, "loss": 1.3184, "step": 9553 }, { "epoch": 0.94, "learning_rate": 1.7692412569852877e-05, "loss": 1.3779, "step": 9554 }, { "epoch": 0.94, "learning_rate": 1.763266584256207e-05, "loss": 1.3828, "step": 9555 }, { "epoch": 0.94, "learning_rate": 1.757301927006316e-05, "loss": 1.5205, "step": 9556 }, { "epoch": 0.94, "learning_rate": 1.7513472858437407e-05, "loss": 1.5059, "step": 9557 }, { "epoch": 0.94, "learning_rate": 1.7454026613755724e-05, "loss": 1.3594, "step": 9558 }, { "epoch": 0.94, "learning_rate": 1.739468054207871e-05, "loss": 1.3223, "step": 9559 }, { "epoch": 0.94, "learning_rate": 1.7335434649456972e-05, "loss": 1.458, "step": 9560 }, { "epoch": 0.94, "learning_rate": 1.7276288941930453e-05, "loss": 1.3184, "step": 9561 }, { "epoch": 0.94, "learning_rate": 1.721724342552966e-05, "loss": 1.2778, "step": 9562 }, { "epoch": 0.94, "learning_rate": 1.7158298106274228e-05, "loss": 1.3047, "step": 9563 }, { "epoch": 0.94, "learning_rate": 1.7099452990173682e-05, "loss": 1.3516, "step": 9564 }, { "epoch": 0.94, "learning_rate": 1.7040708083227553e-05, "loss": 1.3887, "step": 9565 }, { "epoch": 0.94, "learning_rate": 1.698206339142494e-05, "loss": 1.4224, "step": 9566 }, { "epoch": 0.94, "learning_rate": 1.692351892074495e-05, "loss": 1.4326, "step": 9567 }, { "epoch": 0.94, "learning_rate": 1.6865074677156256e-05, "loss": 1.3647, "step": 9568 }, { "epoch": 0.94, "learning_rate": 1.680673066661731e-05, "loss": 1.5244, "step": 9569 }, { "epoch": 0.94, "learning_rate": 1.6748486895076577e-05, "loss": 1.4248, "step": 9570 }, { "epoch": 0.94, "learning_rate": 1.6690343368472084e-05, "loss": 1.2852, "step": 9571 }, { "epoch": 0.94, "learning_rate": 1.6632300092731646e-05, "loss": 1.4482, "step": 9572 }, { "epoch": 0.94, "learning_rate": 1.6574357073773084e-05, "loss": 1.4199, "step": 9573 }, { "epoch": 0.94, "learning_rate": 1.651651431750367e-05, "loss": 1.2725, "step": 9574 }, { "epoch": 0.94, "learning_rate": 1.6458771829820695e-05, "loss": 1.5557, "step": 9575 }, { "epoch": 0.94, "learning_rate": 1.640112961661111e-05, "loss": 1.124, "step": 9576 }, { "epoch": 0.94, "learning_rate": 1.6343587683751658e-05, "loss": 1.3057, "step": 9577 }, { "epoch": 0.94, "learning_rate": 1.6286146037109095e-05, "loss": 1.3843, "step": 9578 }, { "epoch": 0.94, "learning_rate": 1.62288046825394e-05, "loss": 1.4463, "step": 9579 }, { "epoch": 0.94, "learning_rate": 1.6171563625888896e-05, "loss": 1.5029, "step": 9580 }, { "epoch": 0.94, "learning_rate": 1.6114422872993473e-05, "loss": 1.3662, "step": 9581 }, { "epoch": 0.94, "learning_rate": 1.605738242967858e-05, "loss": 1.3818, "step": 9582 }, { "epoch": 0.94, "learning_rate": 1.600044230175979e-05, "loss": 1.2539, "step": 9583 }, { "epoch": 0.94, "learning_rate": 1.594360249504234e-05, "loss": 1.3691, "step": 9584 }, { "epoch": 0.94, "learning_rate": 1.588686301532105e-05, "loss": 1.3936, "step": 9585 }, { "epoch": 0.94, "learning_rate": 1.5830223868380623e-05, "loss": 1.3037, "step": 9586 }, { "epoch": 0.95, "learning_rate": 1.5773685059995767e-05, "loss": 1.3779, "step": 9587 }, { "epoch": 0.95, "learning_rate": 1.5717246595930547e-05, "loss": 1.2832, "step": 9588 }, { "epoch": 0.95, "learning_rate": 1.5660908481939127e-05, "loss": 1.2852, "step": 9589 }, { "epoch": 0.95, "learning_rate": 1.5604670723765257e-05, "loss": 1.4531, "step": 9590 }, { "epoch": 0.95, "learning_rate": 1.5548533327142456e-05, "loss": 1.3477, "step": 9591 }, { "epoch": 0.95, "learning_rate": 1.5492496297794257e-05, "loss": 1.4165, "step": 9592 }, { "epoch": 0.95, "learning_rate": 1.5436559641433645e-05, "loss": 1.4062, "step": 9593 }, { "epoch": 0.95, "learning_rate": 1.5380723363763506e-05, "loss": 1.4141, "step": 9594 }, { "epoch": 0.95, "learning_rate": 1.5324987470476503e-05, "loss": 1.3569, "step": 9595 }, { "epoch": 0.95, "learning_rate": 1.5269351967254986e-05, "loss": 1.2832, "step": 9596 }, { "epoch": 0.95, "learning_rate": 1.521381685977119e-05, "loss": 1.4658, "step": 9597 }, { "epoch": 0.95, "learning_rate": 1.5158382153687145e-05, "loss": 1.2988, "step": 9598 }, { "epoch": 0.95, "learning_rate": 1.5103047854654439e-05, "loss": 1.377, "step": 9599 }, { "epoch": 0.95, "learning_rate": 1.5047813968314561e-05, "loss": 1.3398, "step": 9600 }, { "epoch": 0.95, "learning_rate": 1.4992680500298894e-05, "loss": 1.3408, "step": 9601 }, { "epoch": 0.95, "learning_rate": 1.4937647456228165e-05, "loss": 1.4434, "step": 9602 }, { "epoch": 0.95, "learning_rate": 1.488271484171333e-05, "loss": 1.4385, "step": 9603 }, { "epoch": 0.95, "learning_rate": 1.4827882662354797e-05, "loss": 1.3975, "step": 9604 }, { "epoch": 0.95, "learning_rate": 1.4773150923742873e-05, "loss": 1.373, "step": 9605 }, { "epoch": 0.95, "learning_rate": 1.471851963145765e-05, "loss": 1.2695, "step": 9606 }, { "epoch": 0.95, "learning_rate": 1.4663988791068784e-05, "loss": 1.4482, "step": 9607 }, { "epoch": 0.95, "learning_rate": 1.4609558408136048e-05, "loss": 1.3467, "step": 9608 }, { "epoch": 0.95, "learning_rate": 1.455522848820856e-05, "loss": 1.3257, "step": 9609 }, { "epoch": 0.95, "learning_rate": 1.4500999036825556e-05, "loss": 1.3662, "step": 9610 }, { "epoch": 0.95, "learning_rate": 1.4446870059515615e-05, "loss": 1.4321, "step": 9611 }, { "epoch": 0.95, "learning_rate": 1.4392841561797542e-05, "loss": 1.3408, "step": 9612 }, { "epoch": 0.95, "learning_rate": 1.433891354917971e-05, "loss": 1.3438, "step": 9613 }, { "epoch": 0.95, "learning_rate": 1.4285086027159944e-05, "loss": 1.2563, "step": 9614 }, { "epoch": 0.95, "learning_rate": 1.4231359001226407e-05, "loss": 1.3296, "step": 9615 }, { "epoch": 0.95, "learning_rate": 1.4177732476856497e-05, "loss": 1.4424, "step": 9616 }, { "epoch": 0.95, "learning_rate": 1.412420645951762e-05, "loss": 1.5508, "step": 9617 }, { "epoch": 0.95, "learning_rate": 1.4070780954666962e-05, "loss": 1.4277, "step": 9618 }, { "epoch": 0.95, "learning_rate": 1.4017455967751281e-05, "loss": 1.3408, "step": 9619 }, { "epoch": 0.95, "learning_rate": 1.3964231504207225e-05, "loss": 1.4854, "step": 9620 }, { "epoch": 0.95, "learning_rate": 1.3911107569461234e-05, "loss": 1.3867, "step": 9621 }, { "epoch": 0.95, "learning_rate": 1.3858084168929308e-05, "loss": 1.3818, "step": 9622 }, { "epoch": 0.95, "learning_rate": 1.3805161308017344e-05, "loss": 1.3193, "step": 9623 }, { "epoch": 0.95, "learning_rate": 1.3752338992121027e-05, "loss": 1.3379, "step": 9624 }, { "epoch": 0.95, "learning_rate": 1.3699617226625716e-05, "loss": 1.3242, "step": 9625 }, { "epoch": 0.95, "learning_rate": 1.3646996016906443e-05, "loss": 1.3867, "step": 9626 }, { "epoch": 0.95, "learning_rate": 1.359447536832814e-05, "loss": 1.3213, "step": 9627 }, { "epoch": 0.95, "learning_rate": 1.3542055286245414e-05, "loss": 1.4209, "step": 9628 }, { "epoch": 0.95, "learning_rate": 1.3489735776002543e-05, "loss": 1.4092, "step": 9629 }, { "epoch": 0.95, "learning_rate": 1.3437516842933706e-05, "loss": 1.4336, "step": 9630 }, { "epoch": 0.95, "learning_rate": 1.3385398492362754e-05, "loss": 1.3984, "step": 9631 }, { "epoch": 0.95, "learning_rate": 1.3333380729603329e-05, "loss": 1.3506, "step": 9632 }, { "epoch": 0.95, "learning_rate": 1.3281463559958628e-05, "loss": 1.3965, "step": 9633 }, { "epoch": 0.95, "learning_rate": 1.3229646988721866e-05, "loss": 1.3916, "step": 9634 }, { "epoch": 0.95, "learning_rate": 1.3177931021175815e-05, "loss": 1.3945, "step": 9635 }, { "epoch": 0.95, "learning_rate": 1.3126315662593035e-05, "loss": 1.416, "step": 9636 }, { "epoch": 0.95, "learning_rate": 1.3074800918235874e-05, "loss": 1.3916, "step": 9637 }, { "epoch": 0.95, "learning_rate": 1.302338679335635e-05, "loss": 1.5078, "step": 9638 }, { "epoch": 0.95, "learning_rate": 1.2972073293196384e-05, "loss": 1.4268, "step": 9639 }, { "epoch": 0.95, "learning_rate": 1.2920860422987347e-05, "loss": 1.3721, "step": 9640 }, { "epoch": 0.95, "learning_rate": 1.2869748187950613e-05, "loss": 1.5938, "step": 9641 }, { "epoch": 0.95, "learning_rate": 1.2818736593297131e-05, "loss": 1.3506, "step": 9642 }, { "epoch": 0.95, "learning_rate": 1.2767825644227737e-05, "loss": 1.4541, "step": 9643 }, { "epoch": 0.95, "learning_rate": 1.2717015345932948e-05, "loss": 1.4219, "step": 9644 }, { "epoch": 0.95, "learning_rate": 1.266630570359284e-05, "loss": 1.2554, "step": 9645 }, { "epoch": 0.95, "learning_rate": 1.2615696722377611e-05, "loss": 1.3994, "step": 9646 }, { "epoch": 0.95, "learning_rate": 1.2565188407446693e-05, "loss": 1.3643, "step": 9647 }, { "epoch": 0.95, "learning_rate": 1.2514780763949852e-05, "loss": 1.3984, "step": 9648 }, { "epoch": 0.95, "learning_rate": 1.2464473797025977e-05, "loss": 1.2207, "step": 9649 }, { "epoch": 0.95, "learning_rate": 1.24142675118043e-05, "loss": 1.5, "step": 9650 }, { "epoch": 0.95, "learning_rate": 1.2364161913403172e-05, "loss": 1.4277, "step": 9651 }, { "epoch": 0.95, "learning_rate": 1.2314157006931281e-05, "loss": 1.4424, "step": 9652 }, { "epoch": 0.95, "learning_rate": 1.2264252797486442e-05, "loss": 1.335, "step": 9653 }, { "epoch": 0.95, "learning_rate": 1.2214449290156693e-05, "loss": 1.3599, "step": 9654 }, { "epoch": 0.95, "learning_rate": 1.216474649001964e-05, "loss": 1.3501, "step": 9655 }, { "epoch": 0.95, "learning_rate": 1.2115144402142563e-05, "loss": 1.3193, "step": 9656 }, { "epoch": 0.95, "learning_rate": 1.206564303158253e-05, "loss": 1.458, "step": 9657 }, { "epoch": 0.95, "learning_rate": 1.2016242383386167e-05, "loss": 1.3271, "step": 9658 }, { "epoch": 0.95, "learning_rate": 1.1966942462590336e-05, "loss": 1.335, "step": 9659 }, { "epoch": 0.95, "learning_rate": 1.1917743274221015e-05, "loss": 1.4619, "step": 9660 }, { "epoch": 0.95, "learning_rate": 1.1868644823294194e-05, "loss": 1.4521, "step": 9661 }, { "epoch": 0.95, "learning_rate": 1.1819647114815756e-05, "loss": 1.3848, "step": 9662 }, { "epoch": 0.95, "learning_rate": 1.177075015378093e-05, "loss": 1.3027, "step": 9663 }, { "epoch": 0.95, "learning_rate": 1.1721953945175057e-05, "loss": 1.3281, "step": 9664 }, { "epoch": 0.95, "learning_rate": 1.1673258493972827e-05, "loss": 1.5713, "step": 9665 }, { "epoch": 0.95, "learning_rate": 1.1624663805139158e-05, "loss": 1.4131, "step": 9666 }, { "epoch": 0.95, "learning_rate": 1.1576169883628196e-05, "loss": 1.3857, "step": 9667 }, { "epoch": 0.95, "learning_rate": 1.1527776734384099e-05, "loss": 1.4092, "step": 9668 }, { "epoch": 0.95, "learning_rate": 1.1479484362340475e-05, "loss": 1.3428, "step": 9669 }, { "epoch": 0.95, "learning_rate": 1.1431292772421164e-05, "loss": 1.4561, "step": 9670 }, { "epoch": 0.95, "learning_rate": 1.1383201969539125e-05, "loss": 1.3828, "step": 9671 }, { "epoch": 0.95, "learning_rate": 1.1335211958597547e-05, "loss": 1.4326, "step": 9672 }, { "epoch": 0.95, "learning_rate": 1.128732274448907e-05, "loss": 1.3643, "step": 9673 }, { "epoch": 0.95, "learning_rate": 1.1239534332096014e-05, "loss": 1.3291, "step": 9674 }, { "epoch": 0.95, "learning_rate": 1.11918467262907e-05, "loss": 1.3428, "step": 9675 }, { "epoch": 0.95, "learning_rate": 1.1144259931934797e-05, "loss": 1.3203, "step": 9676 }, { "epoch": 0.95, "learning_rate": 1.109677395388009e-05, "loss": 1.4072, "step": 9677 }, { "epoch": 0.95, "learning_rate": 1.1049388796967818e-05, "loss": 1.3594, "step": 9678 }, { "epoch": 0.95, "learning_rate": 1.1002104466029007e-05, "loss": 1.3213, "step": 9679 }, { "epoch": 0.95, "learning_rate": 1.0954920965884352e-05, "loss": 1.373, "step": 9680 }, { "epoch": 0.95, "learning_rate": 1.0907838301344453e-05, "loss": 1.3994, "step": 9681 }, { "epoch": 0.95, "learning_rate": 1.0860856477209468e-05, "loss": 1.4482, "step": 9682 }, { "epoch": 0.95, "learning_rate": 1.0813975498269235e-05, "loss": 1.4258, "step": 9683 }, { "epoch": 0.95, "learning_rate": 1.0767195369303374e-05, "loss": 1.4648, "step": 9684 }, { "epoch": 0.95, "learning_rate": 1.07205160950814e-05, "loss": 1.3643, "step": 9685 }, { "epoch": 0.95, "learning_rate": 1.0673937680362289e-05, "loss": 1.3364, "step": 9686 }, { "epoch": 0.95, "learning_rate": 1.0627460129894795e-05, "loss": 1.3877, "step": 9687 }, { "epoch": 0.96, "learning_rate": 1.058108344841735e-05, "loss": 1.334, "step": 9688 }, { "epoch": 0.96, "learning_rate": 1.0534807640658395e-05, "loss": 1.4219, "step": 9689 }, { "epoch": 0.96, "learning_rate": 1.04886327113356e-05, "loss": 1.2734, "step": 9690 }, { "epoch": 0.96, "learning_rate": 1.0442558665156865e-05, "loss": 1.3936, "step": 9691 }, { "epoch": 0.96, "learning_rate": 1.0396585506819433e-05, "loss": 1.4355, "step": 9692 }, { "epoch": 0.96, "learning_rate": 1.0350713241010335e-05, "loss": 1.3828, "step": 9693 }, { "epoch": 0.96, "learning_rate": 1.030494187240638e-05, "loss": 1.4189, "step": 9694 }, { "epoch": 0.96, "learning_rate": 1.0259271405674176e-05, "loss": 1.3271, "step": 9695 }, { "epoch": 0.96, "learning_rate": 1.021370184546977e-05, "loss": 1.3931, "step": 9696 }, { "epoch": 0.96, "learning_rate": 1.0168233196439225e-05, "loss": 1.3623, "step": 9697 }, { "epoch": 0.96, "learning_rate": 1.0122865463218167e-05, "loss": 1.4609, "step": 9698 }, { "epoch": 0.96, "learning_rate": 1.0077598650431897e-05, "loss": 1.3799, "step": 9699 }, { "epoch": 0.96, "learning_rate": 1.003243276269561e-05, "loss": 1.4092, "step": 9700 }, { "epoch": 0.96, "learning_rate": 9.987367804613846e-06, "loss": 1.3623, "step": 9701 }, { "epoch": 0.96, "learning_rate": 9.942403780781372e-06, "loss": 1.5146, "step": 9702 }, { "epoch": 0.96, "learning_rate": 9.897540695782193e-06, "loss": 1.4536, "step": 9703 }, { "epoch": 0.96, "learning_rate": 9.852778554190201e-06, "loss": 1.3599, "step": 9704 }, { "epoch": 0.96, "learning_rate": 9.80811736056908e-06, "loss": 1.4727, "step": 9705 }, { "epoch": 0.96, "learning_rate": 9.763557119472188e-06, "loss": 1.373, "step": 9706 }, { "epoch": 0.96, "learning_rate": 9.719097835442448e-06, "loss": 1.4824, "step": 9707 }, { "epoch": 0.96, "learning_rate": 9.674739513012564e-06, "loss": 1.2773, "step": 9708 }, { "epoch": 0.96, "learning_rate": 9.630482156705145e-06, "loss": 1.3965, "step": 9709 }, { "epoch": 0.96, "learning_rate": 9.586325771032245e-06, "loss": 1.4541, "step": 9710 }, { "epoch": 0.96, "learning_rate": 9.54227036049571e-06, "loss": 1.4131, "step": 9711 }, { "epoch": 0.96, "learning_rate": 9.498315929587165e-06, "loss": 1.3643, "step": 9712 }, { "epoch": 0.96, "learning_rate": 9.454462482787696e-06, "loss": 1.3779, "step": 9713 }, { "epoch": 0.96, "learning_rate": 9.410710024568503e-06, "loss": 1.4062, "step": 9714 }, { "epoch": 0.96, "learning_rate": 9.367058559390019e-06, "loss": 1.3574, "step": 9715 }, { "epoch": 0.96, "learning_rate": 9.323508091702904e-06, "loss": 1.4229, "step": 9716 }, { "epoch": 0.96, "learning_rate": 9.280058625947053e-06, "loss": 1.4766, "step": 9717 }, { "epoch": 0.96, "learning_rate": 9.236710166552142e-06, "loss": 1.457, "step": 9718 }, { "epoch": 0.96, "learning_rate": 9.193462717937972e-06, "loss": 1.4482, "step": 9719 }, { "epoch": 0.96, "learning_rate": 9.150316284513461e-06, "loss": 1.3115, "step": 9720 }, { "epoch": 0.96, "learning_rate": 9.107270870677531e-06, "loss": 1.3809, "step": 9721 }, { "epoch": 0.96, "learning_rate": 9.064326480818896e-06, "loss": 1.377, "step": 9722 }, { "epoch": 0.96, "learning_rate": 9.02148311931572e-06, "loss": 1.3672, "step": 9723 }, { "epoch": 0.96, "learning_rate": 8.978740790536067e-06, "loss": 1.3379, "step": 9724 }, { "epoch": 0.96, "learning_rate": 8.936099498837558e-06, "loss": 1.3672, "step": 9725 }, { "epoch": 0.96, "learning_rate": 8.893559248567607e-06, "loss": 1.3076, "step": 9726 }, { "epoch": 0.96, "learning_rate": 8.851120044063298e-06, "loss": 1.5049, "step": 9727 }, { "epoch": 0.96, "learning_rate": 8.808781889651507e-06, "loss": 1.2539, "step": 9728 }, { "epoch": 0.96, "learning_rate": 8.766544789648556e-06, "loss": 1.3936, "step": 9729 }, { "epoch": 0.96, "learning_rate": 8.724408748360557e-06, "loss": 1.3633, "step": 9730 }, { "epoch": 0.96, "learning_rate": 8.682373770083629e-06, "loss": 1.4619, "step": 9731 }, { "epoch": 0.96, "learning_rate": 8.640439859103232e-06, "loss": 1.3779, "step": 9732 }, { "epoch": 0.96, "learning_rate": 8.598607019694726e-06, "loss": 1.2764, "step": 9733 }, { "epoch": 0.96, "learning_rate": 8.55687525612281e-06, "loss": 1.3545, "step": 9734 }, { "epoch": 0.96, "learning_rate": 8.51524457264241e-06, "loss": 1.3838, "step": 9735 }, { "epoch": 0.96, "learning_rate": 8.473714973497804e-06, "loss": 1.3799, "step": 9736 }, { "epoch": 0.96, "learning_rate": 8.432286462922934e-06, "loss": 1.4209, "step": 9737 }, { "epoch": 0.96, "learning_rate": 8.390959045141644e-06, "loss": 1.375, "step": 9738 }, { "epoch": 0.96, "learning_rate": 8.349732724367343e-06, "loss": 1.4424, "step": 9739 }, { "epoch": 0.96, "learning_rate": 8.308607504803223e-06, "loss": 1.4082, "step": 9740 }, { "epoch": 0.96, "learning_rate": 8.267583390641931e-06, "loss": 1.4629, "step": 9741 }, { "epoch": 0.96, "learning_rate": 8.22666038606612e-06, "loss": 1.4102, "step": 9742 }, { "epoch": 0.96, "learning_rate": 8.18583849524801e-06, "loss": 1.3604, "step": 9743 }, { "epoch": 0.96, "learning_rate": 8.145117722349381e-06, "loss": 1.4189, "step": 9744 }, { "epoch": 0.96, "learning_rate": 8.104498071521805e-06, "loss": 1.3643, "step": 9745 }, { "epoch": 0.96, "learning_rate": 8.063979546906741e-06, "loss": 1.3066, "step": 9746 }, { "epoch": 0.96, "learning_rate": 8.02356215263489e-06, "loss": 1.3223, "step": 9747 }, { "epoch": 0.96, "learning_rate": 7.983245892827174e-06, "loss": 1.3848, "step": 9748 }, { "epoch": 0.96, "learning_rate": 7.94303077159375e-06, "loss": 1.3564, "step": 9749 }, { "epoch": 0.96, "learning_rate": 7.902916793034787e-06, "loss": 1.4424, "step": 9750 }, { "epoch": 0.96, "learning_rate": 7.862903961239787e-06, "loss": 1.4014, "step": 9751 }, { "epoch": 0.96, "learning_rate": 7.822992280288489e-06, "loss": 1.4404, "step": 9752 }, { "epoch": 0.96, "learning_rate": 7.783181754249636e-06, "loss": 1.3823, "step": 9753 }, { "epoch": 0.96, "learning_rate": 7.743472387182316e-06, "loss": 1.3325, "step": 9754 }, { "epoch": 0.96, "learning_rate": 7.703864183134734e-06, "loss": 1.4521, "step": 9755 }, { "epoch": 0.96, "learning_rate": 7.664357146145218e-06, "loss": 1.5322, "step": 9756 }, { "epoch": 0.96, "learning_rate": 7.624951280241654e-06, "loss": 1.3691, "step": 9757 }, { "epoch": 0.96, "learning_rate": 7.5856465894413865e-06, "loss": 1.335, "step": 9758 }, { "epoch": 0.96, "learning_rate": 7.5464430777516525e-06, "loss": 1.3955, "step": 9759 }, { "epoch": 0.96, "learning_rate": 7.507340749169478e-06, "loss": 1.5195, "step": 9760 }, { "epoch": 0.96, "learning_rate": 7.46833960768134e-06, "loss": 1.3271, "step": 9761 }, { "epoch": 0.96, "learning_rate": 7.429439657263615e-06, "loss": 1.3193, "step": 9762 }, { "epoch": 0.96, "learning_rate": 7.390640901882017e-06, "loss": 1.3408, "step": 9763 }, { "epoch": 0.96, "learning_rate": 7.351943345492495e-06, "loss": 1.4141, "step": 9764 }, { "epoch": 0.96, "learning_rate": 7.313346992040116e-06, "loss": 1.3945, "step": 9765 }, { "epoch": 0.96, "learning_rate": 7.27485184545984e-06, "loss": 1.3037, "step": 9766 }, { "epoch": 0.96, "learning_rate": 7.236457909676641e-06, "loss": 1.4062, "step": 9767 }, { "epoch": 0.96, "learning_rate": 7.1981651886045e-06, "loss": 1.584, "step": 9768 }, { "epoch": 0.96, "learning_rate": 7.159973686147847e-06, "loss": 1.501, "step": 9769 }, { "epoch": 0.96, "learning_rate": 7.121883406200014e-06, "loss": 1.4053, "step": 9770 }, { "epoch": 0.96, "learning_rate": 7.083894352644671e-06, "loss": 1.3408, "step": 9771 }, { "epoch": 0.96, "learning_rate": 7.046006529354832e-06, "loss": 1.3057, "step": 9772 }, { "epoch": 0.96, "learning_rate": 7.008219940193295e-06, "loss": 1.2617, "step": 9773 }, { "epoch": 0.96, "learning_rate": 6.970534589012312e-06, "loss": 1.4023, "step": 9774 }, { "epoch": 0.96, "learning_rate": 6.932950479654143e-06, "loss": 1.4541, "step": 9775 }, { "epoch": 0.96, "learning_rate": 6.895467615950613e-06, "loss": 1.5244, "step": 9776 }, { "epoch": 0.96, "learning_rate": 6.858086001723107e-06, "loss": 1.4023, "step": 9777 }, { "epoch": 0.96, "learning_rate": 6.82080564078269e-06, "loss": 1.6016, "step": 9778 }, { "epoch": 0.96, "learning_rate": 6.783626536930543e-06, "loss": 1.2925, "step": 9779 }, { "epoch": 0.96, "learning_rate": 6.746548693956744e-06, "loss": 1.3433, "step": 9780 }, { "epoch": 0.96, "learning_rate": 6.709572115641715e-06, "loss": 1.375, "step": 9781 }, { "epoch": 0.96, "learning_rate": 6.672696805755219e-06, "loss": 1.4492, "step": 9782 }, { "epoch": 0.96, "learning_rate": 6.635922768056801e-06, "loss": 1.2529, "step": 9783 }, { "epoch": 0.96, "learning_rate": 6.599250006295687e-06, "loss": 1.335, "step": 9784 }, { "epoch": 0.96, "learning_rate": 6.562678524210774e-06, "loss": 1.332, "step": 9785 }, { "epoch": 0.96, "learning_rate": 6.5262083255306355e-06, "loss": 1.457, "step": 9786 }, { "epoch": 0.96, "learning_rate": 6.489839413973408e-06, "loss": 1.416, "step": 9787 }, { "epoch": 0.96, "learning_rate": 6.453571793247015e-06, "loss": 1.2754, "step": 9788 }, { "epoch": 0.97, "learning_rate": 6.4174054670490536e-06, "loss": 1.1729, "step": 9789 }, { "epoch": 0.97, "learning_rate": 6.381340439066796e-06, "loss": 1.3291, "step": 9790 }, { "epoch": 0.97, "learning_rate": 6.345376712977191e-06, "loss": 1.5186, "step": 9791 }, { "epoch": 0.97, "learning_rate": 6.30951429244675e-06, "loss": 1.3906, "step": 9792 }, { "epoch": 0.97, "learning_rate": 6.27375318113177e-06, "loss": 1.377, "step": 9793 }, { "epoch": 0.97, "learning_rate": 6.238093382678111e-06, "loss": 1.333, "step": 9794 }, { "epoch": 0.97, "learning_rate": 6.202534900721424e-06, "loss": 1.3799, "step": 9795 }, { "epoch": 0.97, "learning_rate": 6.167077738887139e-06, "loss": 1.5088, "step": 9796 }, { "epoch": 0.97, "learning_rate": 6.131721900789921e-06, "loss": 1.3506, "step": 9797 }, { "epoch": 0.97, "learning_rate": 6.096467390034554e-06, "loss": 1.4287, "step": 9798 }, { "epoch": 0.97, "learning_rate": 6.061314210215385e-06, "loss": 1.2319, "step": 9799 }, { "epoch": 0.97, "learning_rate": 6.026262364916213e-06, "loss": 1.335, "step": 9800 }, { "epoch": 0.97, "learning_rate": 5.991311857710846e-06, "loss": 1.4727, "step": 9801 }, { "epoch": 0.97, "learning_rate": 5.956462692162323e-06, "loss": 1.4424, "step": 9802 }, { "epoch": 0.97, "learning_rate": 5.921714871823802e-06, "loss": 1.3975, "step": 9803 }, { "epoch": 0.97, "learning_rate": 5.887068400237894e-06, "loss": 1.3779, "step": 9804 }, { "epoch": 0.97, "learning_rate": 5.852523280936773e-06, "loss": 1.3408, "step": 9805 }, { "epoch": 0.97, "learning_rate": 5.818079517442509e-06, "loss": 1.376, "step": 9806 }, { "epoch": 0.97, "learning_rate": 5.7837371132667405e-06, "loss": 1.3008, "step": 9807 }, { "epoch": 0.97, "learning_rate": 5.749496071910665e-06, "loss": 1.293, "step": 9808 }, { "epoch": 0.97, "learning_rate": 5.715356396865268e-06, "loss": 1.4092, "step": 9809 }, { "epoch": 0.97, "learning_rate": 5.6813180916112095e-06, "loss": 1.3525, "step": 9810 }, { "epoch": 0.97, "learning_rate": 5.647381159618825e-06, "loss": 1.4453, "step": 9811 }, { "epoch": 0.97, "learning_rate": 5.613545604347903e-06, "loss": 1.3496, "step": 9812 }, { "epoch": 0.97, "learning_rate": 5.5798114292482384e-06, "loss": 1.3701, "step": 9813 }, { "epoch": 0.97, "learning_rate": 5.54617863775908e-06, "loss": 1.3193, "step": 9814 }, { "epoch": 0.97, "learning_rate": 5.5126472333093535e-06, "loss": 1.3652, "step": 9815 }, { "epoch": 0.97, "learning_rate": 5.479217219317545e-06, "loss": 1.4395, "step": 9816 }, { "epoch": 0.97, "learning_rate": 5.44588859919215e-06, "loss": 1.2598, "step": 9817 }, { "epoch": 0.97, "learning_rate": 5.412661376330896e-06, "loss": 1.4097, "step": 9818 }, { "epoch": 0.97, "learning_rate": 5.379535554121629e-06, "loss": 1.3477, "step": 9819 }, { "epoch": 0.97, "learning_rate": 5.346511135941312e-06, "loss": 1.3989, "step": 9820 }, { "epoch": 0.97, "learning_rate": 5.313588125157143e-06, "loss": 1.3623, "step": 9821 }, { "epoch": 0.97, "learning_rate": 5.280766525125435e-06, "loss": 1.4395, "step": 9822 }, { "epoch": 0.97, "learning_rate": 5.248046339192625e-06, "loss": 1.4316, "step": 9823 }, { "epoch": 0.97, "learning_rate": 5.2154275706945974e-06, "loss": 1.3701, "step": 9824 }, { "epoch": 0.97, "learning_rate": 5.182910222956916e-06, "loss": 1.3472, "step": 9825 }, { "epoch": 0.97, "learning_rate": 5.1504942992947055e-06, "loss": 1.3682, "step": 9826 }, { "epoch": 0.97, "learning_rate": 5.11817980301299e-06, "loss": 1.4424, "step": 9827 }, { "epoch": 0.97, "learning_rate": 5.085966737406134e-06, "loss": 1.4678, "step": 9828 }, { "epoch": 0.97, "learning_rate": 5.053855105758509e-06, "loss": 1.4404, "step": 9829 }, { "epoch": 0.97, "learning_rate": 5.021844911343943e-06, "loss": 1.3809, "step": 9830 }, { "epoch": 0.97, "learning_rate": 4.989936157425934e-06, "loss": 1.3281, "step": 9831 }, { "epoch": 0.97, "learning_rate": 4.958128847257548e-06, "loss": 1.2754, "step": 9832 }, { "epoch": 0.97, "learning_rate": 4.926422984081858e-06, "loss": 1.3613, "step": 9833 }, { "epoch": 0.97, "learning_rate": 4.8948185711310545e-06, "loss": 1.3408, "step": 9834 }, { "epoch": 0.97, "learning_rate": 4.863315611627561e-06, "loss": 1.5176, "step": 9835 }, { "epoch": 0.97, "learning_rate": 4.831914108783031e-06, "loss": 1.3086, "step": 9836 }, { "epoch": 0.97, "learning_rate": 4.800614065799014e-06, "loss": 1.4058, "step": 9837 }, { "epoch": 0.97, "learning_rate": 4.769415485866513e-06, "loss": 1.4658, "step": 9838 }, { "epoch": 0.97, "learning_rate": 4.738318372166428e-06, "loss": 1.3853, "step": 9839 }, { "epoch": 0.97, "learning_rate": 4.707322727869112e-06, "loss": 1.4863, "step": 9840 }, { "epoch": 0.97, "learning_rate": 4.676428556134704e-06, "loss": 1.4258, "step": 9841 }, { "epoch": 0.97, "learning_rate": 4.645635860112796e-06, "loss": 1.3467, "step": 9842 }, { "epoch": 0.97, "learning_rate": 4.614944642942986e-06, "loss": 1.3975, "step": 9843 }, { "epoch": 0.97, "learning_rate": 4.584354907754107e-06, "loss": 1.4561, "step": 9844 }, { "epoch": 0.97, "learning_rate": 4.5538666576651066e-06, "loss": 1.4189, "step": 9845 }, { "epoch": 0.97, "learning_rate": 4.523479895784055e-06, "loss": 1.3135, "step": 9846 }, { "epoch": 0.97, "learning_rate": 4.493194625209252e-06, "loss": 1.3037, "step": 9847 }, { "epoch": 0.97, "learning_rate": 4.463010849028004e-06, "loss": 1.4199, "step": 9848 }, { "epoch": 0.97, "learning_rate": 4.432928570317962e-06, "loss": 1.4258, "step": 9849 }, { "epoch": 0.97, "learning_rate": 4.402947792146006e-06, "loss": 1.5078, "step": 9850 }, { "epoch": 0.97, "learning_rate": 4.373068517568579e-06, "loss": 1.2925, "step": 9851 }, { "epoch": 0.97, "learning_rate": 4.343290749632134e-06, "loss": 1.4136, "step": 9852 }, { "epoch": 0.97, "learning_rate": 4.313614491372464e-06, "loss": 1.417, "step": 9853 }, { "epoch": 0.97, "learning_rate": 4.28403974581526e-06, "loss": 1.3672, "step": 9854 }, { "epoch": 0.97, "learning_rate": 4.254566515975666e-06, "loss": 1.5645, "step": 9855 }, { "epoch": 0.97, "learning_rate": 4.225194804858612e-06, "loss": 1.4121, "step": 9856 }, { "epoch": 0.97, "learning_rate": 4.195924615458479e-06, "loss": 1.4053, "step": 9857 }, { "epoch": 0.97, "learning_rate": 4.166755950759549e-06, "loss": 1.3696, "step": 9858 }, { "epoch": 0.97, "learning_rate": 4.137688813735552e-06, "loss": 1.4131, "step": 9859 }, { "epoch": 0.97, "learning_rate": 4.108723207350118e-06, "loss": 1.4209, "step": 9860 }, { "epoch": 0.97, "learning_rate": 4.0798591345562185e-06, "loss": 1.4307, "step": 9861 }, { "epoch": 0.97, "learning_rate": 4.05109659829661e-06, "loss": 1.3545, "step": 9862 }, { "epoch": 0.97, "learning_rate": 4.022435601503838e-06, "loss": 1.3818, "step": 9863 }, { "epoch": 0.97, "learning_rate": 3.993876147099784e-06, "loss": 1.3457, "step": 9864 }, { "epoch": 0.97, "learning_rate": 3.965418237996343e-06, "loss": 1.2949, "step": 9865 }, { "epoch": 0.97, "learning_rate": 3.937061877094639e-06, "loss": 1.2949, "step": 9866 }, { "epoch": 0.97, "learning_rate": 3.9088070672859135e-06, "loss": 1.4014, "step": 9867 }, { "epoch": 0.97, "learning_rate": 3.880653811450751e-06, "loss": 1.3242, "step": 9868 }, { "epoch": 0.97, "learning_rate": 3.852602112459302e-06, "loss": 1.2695, "step": 9869 }, { "epoch": 0.97, "learning_rate": 3.8246519731717225e-06, "loss": 1.3301, "step": 9870 }, { "epoch": 0.97, "learning_rate": 3.7968033964373983e-06, "loss": 1.4082, "step": 9871 }, { "epoch": 0.97, "learning_rate": 3.769056385095726e-06, "loss": 1.2666, "step": 9872 }, { "epoch": 0.97, "learning_rate": 3.7414109419754428e-06, "loss": 1.3652, "step": 9873 }, { "epoch": 0.97, "learning_rate": 3.7138670698950716e-06, "loss": 1.4414, "step": 9874 }, { "epoch": 0.97, "learning_rate": 3.6864247716629216e-06, "loss": 1.2764, "step": 9875 }, { "epoch": 0.97, "learning_rate": 3.6590840500767554e-06, "loss": 1.3809, "step": 9876 }, { "epoch": 0.97, "learning_rate": 3.631844907924009e-06, "loss": 1.3916, "step": 9877 }, { "epoch": 0.97, "learning_rate": 3.6047073479816837e-06, "loss": 1.271, "step": 9878 }, { "epoch": 0.97, "learning_rate": 3.5776713730166775e-06, "loss": 1.4023, "step": 9879 }, { "epoch": 0.97, "learning_rate": 3.55073698578523e-06, "loss": 1.3457, "step": 9880 }, { "epoch": 0.97, "learning_rate": 3.523904189033478e-06, "loss": 1.269, "step": 9881 }, { "epoch": 0.97, "learning_rate": 3.4971729854970104e-06, "loss": 1.5137, "step": 9882 }, { "epoch": 0.97, "learning_rate": 3.470543377901314e-06, "loss": 1.498, "step": 9883 }, { "epoch": 0.97, "learning_rate": 3.4440153689611065e-06, "loss": 1.4629, "step": 9884 }, { "epoch": 0.97, "learning_rate": 3.417588961381113e-06, "loss": 1.3232, "step": 9885 }, { "epoch": 0.97, "learning_rate": 3.3912641578555113e-06, "loss": 1.3647, "step": 9886 }, { "epoch": 0.97, "learning_rate": 3.3650409610682664e-06, "loss": 1.335, "step": 9887 }, { "epoch": 0.97, "learning_rate": 3.338919373692795e-06, "loss": 1.5029, "step": 9888 }, { "epoch": 0.97, "learning_rate": 3.3128993983924107e-06, "loss": 1.3447, "step": 9889 }, { "epoch": 0.97, "learning_rate": 3.2869810378197697e-06, "loss": 1.4629, "step": 9890 }, { "epoch": 0.98, "learning_rate": 3.261164294617314e-06, "loss": 1.2129, "step": 9891 }, { "epoch": 0.98, "learning_rate": 3.2354491714172706e-06, "loss": 1.4912, "step": 9892 }, { "epoch": 0.98, "learning_rate": 3.20983567084121e-06, "loss": 1.3896, "step": 9893 }, { "epoch": 0.98, "learning_rate": 3.184323795500488e-06, "loss": 1.4258, "step": 9894 }, { "epoch": 0.98, "learning_rate": 3.158913547996245e-06, "loss": 1.5088, "step": 9895 }, { "epoch": 0.98, "learning_rate": 3.1336049309190763e-06, "loss": 1.4131, "step": 9896 }, { "epoch": 0.98, "learning_rate": 3.1083979468491397e-06, "loss": 1.2964, "step": 9897 }, { "epoch": 0.98, "learning_rate": 3.0832925983563796e-06, "loss": 1.291, "step": 9898 }, { "epoch": 0.98, "learning_rate": 3.0582888880004154e-06, "loss": 1.25, "step": 9899 }, { "epoch": 0.98, "learning_rate": 3.0333868183304304e-06, "loss": 1.4492, "step": 9900 }, { "epoch": 0.98, "learning_rate": 3.0085863918852817e-06, "loss": 1.3604, "step": 9901 }, { "epoch": 0.98, "learning_rate": 2.983887611193392e-06, "loss": 1.2734, "step": 9902 }, { "epoch": 0.98, "learning_rate": 2.959290478772858e-06, "loss": 1.2705, "step": 9903 }, { "epoch": 0.98, "learning_rate": 2.9347949971313403e-06, "loss": 1.291, "step": 9904 }, { "epoch": 0.98, "learning_rate": 2.910401168766397e-06, "loss": 1.4473, "step": 9905 }, { "epoch": 0.98, "learning_rate": 2.886108996164927e-06, "loss": 1.3701, "step": 9906 }, { "epoch": 0.98, "learning_rate": 2.861918481803505e-06, "loss": 1.3359, "step": 9907 }, { "epoch": 0.98, "learning_rate": 2.8378296281486027e-06, "loss": 1.2241, "step": 9908 }, { "epoch": 0.98, "learning_rate": 2.8138424376560335e-06, "loss": 1.3438, "step": 9909 }, { "epoch": 0.98, "learning_rate": 2.7899569127712853e-06, "loss": 1.3994, "step": 9910 }, { "epoch": 0.98, "learning_rate": 2.7661730559296327e-06, "loss": 1.3975, "step": 9911 }, { "epoch": 0.98, "learning_rate": 2.7424908695559135e-06, "loss": 1.2988, "step": 9912 }, { "epoch": 0.98, "learning_rate": 2.71891035606453e-06, "loss": 1.46, "step": 9913 }, { "epoch": 0.98, "learning_rate": 2.695431517859559e-06, "loss": 1.394, "step": 9914 }, { "epoch": 0.98, "learning_rate": 2.672054357334863e-06, "loss": 1.3809, "step": 9915 }, { "epoch": 0.98, "learning_rate": 2.6487788768736475e-06, "loss": 1.415, "step": 9916 }, { "epoch": 0.98, "learning_rate": 2.6256050788489027e-06, "loss": 1.3506, "step": 9917 }, { "epoch": 0.98, "learning_rate": 2.6025329656234054e-06, "loss": 1.6084, "step": 9918 }, { "epoch": 0.98, "learning_rate": 2.5795625395492738e-06, "loss": 1.3105, "step": 9919 }, { "epoch": 0.98, "learning_rate": 2.556693802968524e-06, "loss": 1.2852, "step": 9920 }, { "epoch": 0.98, "learning_rate": 2.5339267582125124e-06, "loss": 1.4873, "step": 9921 }, { "epoch": 0.98, "learning_rate": 2.5112614076026054e-06, "loss": 1.4531, "step": 9922 }, { "epoch": 0.98, "learning_rate": 2.488697753449398e-06, "loss": 1.3027, "step": 9923 }, { "epoch": 0.98, "learning_rate": 2.4662357980533845e-06, "loss": 1.3003, "step": 9924 }, { "epoch": 0.98, "learning_rate": 2.4438755437046212e-06, "loss": 1.375, "step": 9925 }, { "epoch": 0.98, "learning_rate": 2.4216169926827294e-06, "loss": 1.2744, "step": 9926 }, { "epoch": 0.98, "learning_rate": 2.3994601472572266e-06, "loss": 1.4561, "step": 9927 }, { "epoch": 0.98, "learning_rate": 2.3774050096868617e-06, "loss": 1.4209, "step": 9928 }, { "epoch": 0.98, "learning_rate": 2.3554515822201693e-06, "loss": 1.3564, "step": 9929 }, { "epoch": 0.98, "learning_rate": 2.3335998670955818e-06, "loss": 1.374, "step": 9930 }, { "epoch": 0.98, "learning_rate": 2.3118498665407605e-06, "loss": 1.3525, "step": 9931 }, { "epoch": 0.98, "learning_rate": 2.290201582773266e-06, "loss": 1.4668, "step": 9932 }, { "epoch": 0.98, "learning_rate": 2.2686550180001098e-06, "loss": 1.3447, "step": 9933 }, { "epoch": 0.98, "learning_rate": 2.2472101744179795e-06, "loss": 1.4131, "step": 9934 }, { "epoch": 0.98, "learning_rate": 2.2258670542133484e-06, "loss": 1.3604, "step": 9935 }, { "epoch": 0.98, "learning_rate": 2.2046256595622537e-06, "loss": 1.2061, "step": 9936 }, { "epoch": 0.98, "learning_rate": 2.1834859926301853e-06, "loss": 1.2891, "step": 9937 }, { "epoch": 0.98, "learning_rate": 2.162448055572308e-06, "loss": 1.3623, "step": 9938 }, { "epoch": 0.98, "learning_rate": 2.141511850533684e-06, "loss": 1.333, "step": 9939 }, { "epoch": 0.98, "learning_rate": 2.1206773796487168e-06, "loss": 1.2451, "step": 9940 }, { "epoch": 0.98, "learning_rate": 2.0999446450414848e-06, "loss": 1.4443, "step": 9941 }, { "epoch": 0.98, "learning_rate": 2.0793136488258534e-06, "loss": 1.3516, "step": 9942 }, { "epoch": 0.98, "learning_rate": 2.0587843931051398e-06, "loss": 1.3955, "step": 9943 }, { "epoch": 0.98, "learning_rate": 2.038356879972336e-06, "loss": 1.4688, "step": 9944 }, { "epoch": 0.98, "learning_rate": 2.018031111510221e-06, "loss": 1.3477, "step": 9945 }, { "epoch": 0.98, "learning_rate": 1.997807089790804e-06, "loss": 1.4033, "step": 9946 }, { "epoch": 0.98, "learning_rate": 1.977684816876102e-06, "loss": 1.2397, "step": 9947 }, { "epoch": 0.98, "learning_rate": 1.9576642948176957e-06, "loss": 1.3257, "step": 9948 }, { "epoch": 0.98, "learning_rate": 1.9377455256567313e-06, "loss": 1.3887, "step": 9949 }, { "epoch": 0.98, "learning_rate": 1.9179285114238054e-06, "loss": 1.3564, "step": 9950 }, { "epoch": 0.98, "learning_rate": 1.8982132541394137e-06, "loss": 1.2949, "step": 9951 }, { "epoch": 0.98, "learning_rate": 1.8785997558136147e-06, "loss": 1.3643, "step": 9952 }, { "epoch": 0.98, "learning_rate": 1.8590880184460313e-06, "loss": 1.4375, "step": 9953 }, { "epoch": 0.98, "learning_rate": 1.8396780440259608e-06, "loss": 1.2393, "step": 9954 }, { "epoch": 0.98, "learning_rate": 1.820369834532265e-06, "loss": 1.4336, "step": 9955 }, { "epoch": 0.98, "learning_rate": 1.8011633919333692e-06, "loss": 1.269, "step": 9956 }, { "epoch": 0.98, "learning_rate": 1.7820587181875958e-06, "loss": 1.3164, "step": 9957 }, { "epoch": 0.98, "learning_rate": 1.7630558152426091e-06, "loss": 1.3877, "step": 9958 }, { "epoch": 0.98, "learning_rate": 1.7441546850357482e-06, "loss": 1.3311, "step": 9959 }, { "epoch": 0.98, "learning_rate": 1.7253553294941381e-06, "loss": 1.375, "step": 9960 }, { "epoch": 0.98, "learning_rate": 1.7066577505344682e-06, "loss": 1.3193, "step": 9961 }, { "epoch": 0.98, "learning_rate": 1.6880619500627692e-06, "loss": 1.5576, "step": 9962 }, { "epoch": 0.98, "learning_rate": 1.6695679299751908e-06, "loss": 1.3555, "step": 9963 }, { "epoch": 0.98, "learning_rate": 1.6511756921571141e-06, "loss": 1.3838, "step": 9964 }, { "epoch": 0.98, "learning_rate": 1.6328852384837057e-06, "loss": 1.4648, "step": 9965 }, { "epoch": 0.98, "learning_rate": 1.6146965708196958e-06, "loss": 1.4697, "step": 9966 }, { "epoch": 0.98, "learning_rate": 1.5966096910193795e-06, "loss": 1.3105, "step": 9967 }, { "epoch": 0.98, "learning_rate": 1.5786246009269479e-06, "loss": 1.3594, "step": 9968 }, { "epoch": 0.98, "learning_rate": 1.5607413023759342e-06, "loss": 1.2676, "step": 9969 }, { "epoch": 0.98, "learning_rate": 1.5429597971896581e-06, "loss": 1.4326, "step": 9970 }, { "epoch": 0.98, "learning_rate": 1.5252800871807805e-06, "loss": 1.3525, "step": 9971 }, { "epoch": 0.98, "learning_rate": 1.5077021741520814e-06, "loss": 1.5176, "step": 9972 }, { "epoch": 0.98, "learning_rate": 1.49022605989535e-06, "loss": 1.375, "step": 9973 }, { "epoch": 0.98, "learning_rate": 1.472851746192605e-06, "loss": 1.4229, "step": 9974 }, { "epoch": 0.98, "learning_rate": 1.4555792348150966e-06, "loss": 1.4961, "step": 9975 }, { "epoch": 0.98, "learning_rate": 1.438408527523638e-06, "loss": 1.4678, "step": 9976 }, { "epoch": 0.98, "learning_rate": 1.4213396260691624e-06, "loss": 1.3701, "step": 9977 }, { "epoch": 0.98, "learning_rate": 1.4043725321916113e-06, "loss": 1.4463, "step": 9978 }, { "epoch": 0.98, "learning_rate": 1.3875072476209338e-06, "loss": 1.3779, "step": 9979 }, { "epoch": 0.98, "learning_rate": 1.3707437740765326e-06, "loss": 1.3389, "step": 9980 }, { "epoch": 0.98, "learning_rate": 1.3540821132674853e-06, "loss": 1.459, "step": 9981 }, { "epoch": 0.98, "learning_rate": 1.3375222668925436e-06, "loss": 1.4453, "step": 9982 }, { "epoch": 0.98, "learning_rate": 1.3210642366400238e-06, "loss": 1.3711, "step": 9983 }, { "epoch": 0.98, "learning_rate": 1.3047080241879173e-06, "loss": 1.4717, "step": 9984 }, { "epoch": 0.98, "learning_rate": 1.2884536312035567e-06, "loss": 1.3765, "step": 9985 }, { "epoch": 0.98, "learning_rate": 1.272301059344394e-06, "loss": 1.415, "step": 9986 }, { "epoch": 0.98, "learning_rate": 1.2562503102570011e-06, "loss": 1.3467, "step": 9987 }, { "epoch": 0.98, "learning_rate": 1.2403013855779577e-06, "loss": 1.3086, "step": 9988 }, { "epoch": 0.98, "learning_rate": 1.2244542869331854e-06, "loss": 1.3369, "step": 9989 }, { "epoch": 0.98, "learning_rate": 1.2087090159383918e-06, "loss": 1.4219, "step": 9990 }, { "epoch": 0.98, "learning_rate": 1.1930655741988483e-06, "loss": 1.3779, "step": 9991 }, { "epoch": 0.99, "learning_rate": 1.1775239633093904e-06, "loss": 1.3071, "step": 9992 }, { "epoch": 0.99, "learning_rate": 1.1620841848545282e-06, "loss": 1.377, "step": 9993 }, { "epoch": 0.99, "learning_rate": 1.1467462404084473e-06, "loss": 1.3506, "step": 9994 }, { "epoch": 0.99, "learning_rate": 1.1315101315348963e-06, "loss": 1.4375, "step": 9995 }, { "epoch": 0.99, "learning_rate": 1.1163758597871886e-06, "loss": 1.1514, "step": 9996 }, { "epoch": 0.99, "learning_rate": 1.1013434267083122e-06, "loss": 1.4053, "step": 9997 }, { "epoch": 0.99, "learning_rate": 1.0864128338308188e-06, "loss": 1.4023, "step": 9998 }, { "epoch": 0.99, "learning_rate": 1.0715840826769353e-06, "loss": 1.4062, "step": 9999 }, { "epoch": 0.99, "learning_rate": 1.0568571747585632e-06, "loss": 1.3525, "step": 10000 }, { "epoch": 0.99, "learning_rate": 1.0422321115770571e-06, "loss": 1.498, "step": 10001 }, { "epoch": 0.99, "learning_rate": 1.027708894623558e-06, "loss": 1.248, "step": 10002 }, { "epoch": 0.99, "learning_rate": 1.013287525378659e-06, "loss": 1.4629, "step": 10003 }, { "epoch": 0.99, "learning_rate": 9.989680053127391e-07, "loss": 1.3447, "step": 10004 }, { "epoch": 0.99, "learning_rate": 9.847503358857424e-07, "loss": 1.5332, "step": 10005 }, { "epoch": 0.99, "learning_rate": 9.706345185469535e-07, "loss": 1.2979, "step": 10006 }, { "epoch": 0.99, "learning_rate": 9.566205547357765e-07, "loss": 1.3496, "step": 10007 }, { "epoch": 0.99, "learning_rate": 9.427084458808466e-07, "loss": 1.3994, "step": 10008 }, { "epoch": 0.99, "learning_rate": 9.288981934006957e-07, "loss": 1.2974, "step": 10009 }, { "epoch": 0.99, "learning_rate": 9.151897987029756e-07, "loss": 1.3242, "step": 10010 }, { "epoch": 0.99, "learning_rate": 9.015832631856791e-07, "loss": 1.3877, "step": 10011 }, { "epoch": 0.99, "learning_rate": 8.880785882358078e-07, "loss": 1.2139, "step": 10012 }, { "epoch": 0.99, "learning_rate": 8.746757752301493e-07, "loss": 1.3311, "step": 10013 }, { "epoch": 0.99, "learning_rate": 8.613748255352771e-07, "loss": 1.3604, "step": 10014 }, { "epoch": 0.99, "learning_rate": 8.481757405073287e-07, "loss": 1.3535, "step": 10015 }, { "epoch": 0.99, "learning_rate": 8.350785214917833e-07, "loss": 1.458, "step": 10016 }, { "epoch": 0.99, "learning_rate": 8.220831698240172e-07, "loss": 1.3545, "step": 10017 }, { "epoch": 0.99, "learning_rate": 8.091896868289706e-07, "loss": 1.2695, "step": 10018 }, { "epoch": 0.99, "learning_rate": 7.963980738211474e-07, "loss": 1.4092, "step": 10019 }, { "epoch": 0.99, "learning_rate": 7.837083321046157e-07, "loss": 1.3911, "step": 10020 }, { "epoch": 0.99, "learning_rate": 7.711204629733404e-07, "loss": 1.3198, "step": 10021 }, { "epoch": 0.99, "learning_rate": 7.586344677104062e-07, "loss": 1.3672, "step": 10022 }, { "epoch": 0.99, "learning_rate": 7.46250347589017e-07, "loss": 1.499, "step": 10023 }, { "epoch": 0.99, "learning_rate": 7.339681038716073e-07, "loss": 1.3955, "step": 10024 }, { "epoch": 0.99, "learning_rate": 7.217877378103976e-07, "loss": 1.3848, "step": 10025 }, { "epoch": 0.99, "learning_rate": 7.097092506473945e-07, "loss": 1.3721, "step": 10026 }, { "epoch": 0.99, "learning_rate": 6.977326436138354e-07, "loss": 1.4736, "step": 10027 }, { "epoch": 0.99, "learning_rate": 6.858579179307434e-07, "loss": 1.4414, "step": 10028 }, { "epoch": 0.99, "learning_rate": 6.740850748089278e-07, "loss": 1.3359, "step": 10029 }, { "epoch": 0.99, "learning_rate": 6.624141154485397e-07, "loss": 1.3232, "step": 10030 }, { "epoch": 0.99, "learning_rate": 6.508450410396272e-07, "loss": 1.5244, "step": 10031 }, { "epoch": 0.99, "learning_rate": 6.393778527614691e-07, "loss": 1.2583, "step": 10032 }, { "epoch": 0.99, "learning_rate": 6.280125517832414e-07, "loss": 1.2837, "step": 10033 }, { "epoch": 0.99, "learning_rate": 6.167491392636837e-07, "loss": 1.3394, "step": 10034 }, { "epoch": 0.99, "learning_rate": 6.055876163512108e-07, "loss": 1.4336, "step": 10035 }, { "epoch": 0.99, "learning_rate": 5.945279841836904e-07, "loss": 1.3145, "step": 10036 }, { "epoch": 0.99, "learning_rate": 5.835702438886648e-07, "loss": 1.4561, "step": 10037 }, { "epoch": 0.99, "learning_rate": 5.727143965833514e-07, "loss": 1.2529, "step": 10038 }, { "epoch": 0.99, "learning_rate": 5.619604433744208e-07, "loss": 1.3848, "step": 10039 }, { "epoch": 0.99, "learning_rate": 5.513083853584399e-07, "loss": 1.4678, "step": 10040 }, { "epoch": 0.99, "learning_rate": 5.407582236213182e-07, "loss": 1.4077, "step": 10041 }, { "epoch": 0.99, "learning_rate": 5.303099592386395e-07, "loss": 1.3662, "step": 10042 }, { "epoch": 0.99, "learning_rate": 5.19963593275663e-07, "loss": 1.3623, "step": 10043 }, { "epoch": 0.99, "learning_rate": 5.097191267872115e-07, "loss": 1.4082, "step": 10044 }, { "epoch": 0.99, "learning_rate": 4.995765608177827e-07, "loss": 1.3623, "step": 10045 }, { "epoch": 0.99, "learning_rate": 4.895358964013274e-07, "loss": 1.4121, "step": 10046 }, { "epoch": 0.99, "learning_rate": 4.795971345616934e-07, "loss": 1.3887, "step": 10047 }, { "epoch": 0.99, "learning_rate": 4.6976027631195905e-07, "loss": 1.4229, "step": 10048 }, { "epoch": 0.99, "learning_rate": 4.600253226552109e-07, "loss": 1.4053, "step": 10049 }, { "epoch": 0.99, "learning_rate": 4.503922745838773e-07, "loss": 1.2676, "step": 10050 }, { "epoch": 0.99, "learning_rate": 4.4086113307995057e-07, "loss": 1.3584, "step": 10051 }, { "epoch": 0.99, "learning_rate": 4.314318991153199e-07, "loss": 1.3735, "step": 10052 }, { "epoch": 0.99, "learning_rate": 4.2210457365121637e-07, "loss": 1.4277, "step": 10053 }, { "epoch": 0.99, "learning_rate": 4.12879157638657e-07, "loss": 1.3574, "step": 10054 }, { "epoch": 0.99, "learning_rate": 4.037556520182228e-07, "loss": 1.3779, "step": 10055 }, { "epoch": 0.99, "learning_rate": 3.9473405771994766e-07, "loss": 1.375, "step": 10056 }, { "epoch": 0.99, "learning_rate": 3.858143756636512e-07, "loss": 1.4502, "step": 10057 }, { "epoch": 0.99, "learning_rate": 3.769966067587172e-07, "loss": 1.3779, "step": 10058 }, { "epoch": 0.99, "learning_rate": 3.6828075190420416e-07, "loss": 1.2949, "step": 10059 }, { "epoch": 0.99, "learning_rate": 3.5966681198873475e-07, "loss": 1.3438, "step": 10060 }, { "epoch": 0.99, "learning_rate": 3.511547878903842e-07, "loss": 1.4043, "step": 10061 }, { "epoch": 0.99, "learning_rate": 3.42744680477125e-07, "loss": 1.4121, "step": 10062 }, { "epoch": 0.99, "learning_rate": 3.3443649060627134e-07, "loss": 1.2749, "step": 10063 }, { "epoch": 0.99, "learning_rate": 3.2623021912492336e-07, "loss": 1.3916, "step": 10064 }, { "epoch": 0.99, "learning_rate": 3.1812586686963405e-07, "loss": 1.3999, "step": 10065 }, { "epoch": 0.99, "learning_rate": 3.1012343466685357e-07, "loss": 1.4258, "step": 10066 }, { "epoch": 0.99, "learning_rate": 3.0222292333226263e-07, "loss": 1.333, "step": 10067 }, { "epoch": 0.99, "learning_rate": 2.9442433367143916e-07, "loss": 1.3096, "step": 10068 }, { "epoch": 0.99, "learning_rate": 2.8672766647952486e-07, "loss": 1.3584, "step": 10069 }, { "epoch": 0.99, "learning_rate": 2.7913292254100333e-07, "loss": 1.3848, "step": 10070 }, { "epoch": 0.99, "learning_rate": 2.716401026303661e-07, "loss": 1.3486, "step": 10071 }, { "epoch": 0.99, "learning_rate": 2.6424920751144663e-07, "loss": 1.3438, "step": 10072 }, { "epoch": 0.99, "learning_rate": 2.5696023793786437e-07, "loss": 1.3867, "step": 10073 }, { "epoch": 0.99, "learning_rate": 2.4977319465258053e-07, "loss": 1.4473, "step": 10074 }, { "epoch": 0.99, "learning_rate": 2.426880783884533e-07, "loss": 1.3496, "step": 10075 }, { "epoch": 0.99, "learning_rate": 2.357048898677938e-07, "loss": 1.4141, "step": 10076 }, { "epoch": 0.99, "learning_rate": 2.2882362980258809e-07, "loss": 1.334, "step": 10077 }, { "epoch": 0.99, "learning_rate": 2.2204429889427501e-07, "loss": 1.332, "step": 10078 }, { "epoch": 0.99, "learning_rate": 2.153668978341905e-07, "loss": 1.3672, "step": 10079 }, { "epoch": 0.99, "learning_rate": 2.0879142730301227e-07, "loss": 1.4277, "step": 10080 }, { "epoch": 0.99, "learning_rate": 2.0231788797109295e-07, "loss": 1.3936, "step": 10081 }, { "epoch": 0.99, "learning_rate": 1.9594628049857123e-07, "loss": 1.2744, "step": 10082 }, { "epoch": 0.99, "learning_rate": 1.896766055349275e-07, "loss": 1.4316, "step": 10083 }, { "epoch": 0.99, "learning_rate": 1.835088637194282e-07, "loss": 1.5137, "step": 10084 }, { "epoch": 0.99, "learning_rate": 1.7744305568079267e-07, "loss": 1.3848, "step": 10085 }, { "epoch": 0.99, "learning_rate": 1.7147918203763714e-07, "loss": 1.4639, "step": 10086 }, { "epoch": 0.99, "learning_rate": 1.656172433978087e-07, "loss": 1.5059, "step": 10087 }, { "epoch": 0.99, "learning_rate": 1.5985724035905147e-07, "loss": 1.4316, "step": 10088 }, { "epoch": 0.99, "learning_rate": 1.5419917350845136e-07, "loss": 1.4307, "step": 10089 }, { "epoch": 0.99, "learning_rate": 1.4864304342310231e-07, "loss": 1.4102, "step": 10090 }, { "epoch": 0.99, "learning_rate": 1.4318885066932907e-07, "loss": 1.4404, "step": 10091 }, { "epoch": 0.99, "learning_rate": 1.3783659580313134e-07, "loss": 1.3623, "step": 10092 }, { "epoch": 0.99, "learning_rate": 1.325862793704058e-07, "loss": 1.4824, "step": 10093 }, { "epoch": 1.0, "learning_rate": 1.274379019061689e-07, "loss": 1.3467, "step": 10094 }, { "epoch": 1.0, "learning_rate": 1.223914639355561e-07, "loss": 1.4072, "step": 10095 }, { "epoch": 1.0, "learning_rate": 1.174469659729338e-07, "loss": 1.4082, "step": 10096 }, { "epoch": 1.0, "learning_rate": 1.1260440852234321e-07, "loss": 1.4014, "step": 10097 }, { "epoch": 1.0, "learning_rate": 1.0786379207761154e-07, "loss": 1.3867, "step": 10098 }, { "epoch": 1.0, "learning_rate": 1.032251171221299e-07, "loss": 1.21, "step": 10099 }, { "epoch": 1.0, "learning_rate": 9.868838412852021e-08, "loss": 1.4033, "step": 10100 }, { "epoch": 1.0, "learning_rate": 9.425359355963448e-08, "loss": 1.4102, "step": 10101 }, { "epoch": 1.0, "learning_rate": 8.992074586744447e-08, "loss": 1.3613, "step": 10102 }, { "epoch": 1.0, "learning_rate": 8.568984149370795e-08, "loss": 1.377, "step": 10103 }, { "epoch": 1.0, "learning_rate": 8.156088086985757e-08, "loss": 1.4209, "step": 10104 }, { "epoch": 1.0, "learning_rate": 7.75338644167789e-08, "loss": 1.3887, "step": 10105 }, { "epoch": 1.0, "learning_rate": 7.360879254503238e-08, "loss": 1.4023, "step": 10106 }, { "epoch": 1.0, "learning_rate": 6.978566565485344e-08, "loss": 1.314, "step": 10107 }, { "epoch": 1.0, "learning_rate": 6.606448413593036e-08, "loss": 1.2729, "step": 10108 }, { "epoch": 1.0, "learning_rate": 6.244524836762633e-08, "loss": 1.3311, "step": 10109 }, { "epoch": 1.0, "learning_rate": 5.8927958719090516e-08, "loss": 1.4238, "step": 10110 }, { "epoch": 1.0, "learning_rate": 5.551261554881393e-08, "loss": 1.3525, "step": 10111 }, { "epoch": 1.0, "learning_rate": 5.2199219204962515e-08, "loss": 1.3965, "step": 10112 }, { "epoch": 1.0, "learning_rate": 4.8987770025488156e-08, "loss": 1.3154, "step": 10113 }, { "epoch": 1.0, "learning_rate": 4.587826833757358e-08, "loss": 1.4395, "step": 10114 }, { "epoch": 1.0, "learning_rate": 4.287071445852053e-08, "loss": 1.4941, "step": 10115 }, { "epoch": 1.0, "learning_rate": 3.996510869475056e-08, "loss": 1.4473, "step": 10116 }, { "epoch": 1.0, "learning_rate": 3.7161451342471176e-08, "loss": 1.3755, "step": 10117 }, { "epoch": 1.0, "learning_rate": 3.4459742687675824e-08, "loss": 1.3164, "step": 10118 }, { "epoch": 1.0, "learning_rate": 3.185998300581083e-08, "loss": 1.3945, "step": 10119 }, { "epoch": 1.0, "learning_rate": 2.9362172561775425e-08, "loss": 1.3359, "step": 10120 }, { "epoch": 1.0, "learning_rate": 2.696631161036578e-08, "loss": 1.458, "step": 10121 }, { "epoch": 1.0, "learning_rate": 2.4672400395830963e-08, "loss": 1.5078, "step": 10122 }, { "epoch": 1.0, "learning_rate": 2.2480439151983943e-08, "loss": 1.4023, "step": 10123 }, { "epoch": 1.0, "learning_rate": 2.0390428102201596e-08, "loss": 1.3193, "step": 10124 }, { "epoch": 1.0, "learning_rate": 1.8402367459868785e-08, "loss": 1.4443, "step": 10125 }, { "epoch": 1.0, "learning_rate": 1.6516257427379165e-08, "loss": 1.4766, "step": 10126 }, { "epoch": 1.0, "learning_rate": 1.4732098197134392e-08, "loss": 1.4121, "step": 10127 }, { "epoch": 1.0, "learning_rate": 1.3049889951100013e-08, "loss": 1.3066, "step": 10128 }, { "epoch": 1.0, "learning_rate": 1.1469632860694468e-08, "loss": 1.4092, "step": 10129 }, { "epoch": 1.0, "learning_rate": 9.991327087011115e-09, "loss": 1.4585, "step": 10130 }, { "epoch": 1.0, "learning_rate": 8.614972780818242e-09, "loss": 1.2842, "step": 10131 }, { "epoch": 1.0, "learning_rate": 7.340570082448039e-09, "loss": 1.4526, "step": 10132 }, { "epoch": 1.0, "learning_rate": 6.1681191217966005e-09, "loss": 1.4248, "step": 10133 }, { "epoch": 1.0, "learning_rate": 5.097620018434945e-09, "loss": 1.4756, "step": 10134 }, { "epoch": 1.0, "learning_rate": 4.129072881497997e-09, "loss": 1.3228, "step": 10135 }, { "epoch": 1.0, "learning_rate": 3.2624778096845833e-09, "loss": 1.332, "step": 10136 }, { "epoch": 1.0, "learning_rate": 2.4978348913684557e-09, "loss": 1.3433, "step": 10137 }, { "epoch": 1.0, "learning_rate": 1.8351442044872712e-09, "loss": 1.3618, "step": 10138 }, { "epoch": 1.0, "learning_rate": 1.274405816653612e-09, "loss": 1.4082, "step": 10139 }, { "epoch": 1.0, "learning_rate": 8.156197850439639e-10, "loss": 1.3867, "step": 10140 }, { "epoch": 1.0, "learning_rate": 4.587861563987161e-10, "loss": 1.3926, "step": 10141 }, { "epoch": 1.0, "learning_rate": 2.0390496702216156e-10, "loss": 1.4375, "step": 10142 }, { "epoch": 1.0, "learning_rate": 5.097624311556359e-11, "loss": 1.3105, "step": 10143 }, { "epoch": 1.0, "learning_rate": 0.0, "loss": 1.3711, "step": 10144 }, { "epoch": 1.0, "step": 10144, "total_flos": 2.076177358303684e+19, "train_loss": 1.6492533397975404, "train_runtime": 51144.1577, "train_samples_per_second": 25.386, "train_steps_per_second": 0.198 } ], "max_steps": 10144, "num_train_epochs": 1, "total_flos": 2.076177358303684e+19, "trial_name": null, "trial_params": null }