| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.9990091652216995, | |
| "eval_steps": 500, | |
| "global_step": 4036, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.278688524590164e-08, | |
| "loss": 1.8177, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 6.557377049180328e-08, | |
| "loss": 1.8121, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.836065573770492e-08, | |
| "loss": 1.8325, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.3114754098360656e-07, | |
| "loss": 1.841, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.6393442622950818e-07, | |
| "loss": 1.7706, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 1.9672131147540984e-07, | |
| "loss": 1.8287, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 2.2950819672131146e-07, | |
| "loss": 1.8792, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.622950819672131e-07, | |
| "loss": 1.8158, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 2.950819672131147e-07, | |
| "loss": 1.8152, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.2786885245901637e-07, | |
| "loss": 1.7649, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.60655737704918e-07, | |
| "loss": 1.8348, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 3.9344262295081967e-07, | |
| "loss": 1.8553, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.2622950819672127e-07, | |
| "loss": 1.8001, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.590163934426229e-07, | |
| "loss": 1.8601, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 4.918032786885245e-07, | |
| "loss": 1.8731, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 5.245901639344262e-07, | |
| "loss": 1.8386, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 5.573770491803278e-07, | |
| "loss": 1.8927, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 5.901639344262294e-07, | |
| "loss": 1.8632, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.229508196721311e-07, | |
| "loss": 1.8382, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.557377049180327e-07, | |
| "loss": 1.7967, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 6.885245901639343e-07, | |
| "loss": 1.8523, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 7.21311475409836e-07, | |
| "loss": 1.8736, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 7.540983606557376e-07, | |
| "loss": 1.8387, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 7.868852459016393e-07, | |
| "loss": 1.7413, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.196721311475409e-07, | |
| "loss": 1.8022, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.524590163934425e-07, | |
| "loss": 1.8004, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 8.852459016393443e-07, | |
| "loss": 1.8658, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.180327868852458e-07, | |
| "loss": 1.819, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.508196721311474e-07, | |
| "loss": 1.8115, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 9.83606557377049e-07, | |
| "loss": 1.7732, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.0163934426229509e-06, | |
| "loss": 1.8582, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 1.0491803278688525e-06, | |
| "loss": 1.8034, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.081967213114754e-06, | |
| "loss": 1.7767, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.1147540983606557e-06, | |
| "loss": 1.7388, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.1475409836065573e-06, | |
| "loss": 1.891, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.1803278688524589e-06, | |
| "loss": 1.847, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 1.2131147540983607e-06, | |
| "loss": 1.8508, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.2459016393442623e-06, | |
| "loss": 1.8578, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.2786885245901639e-06, | |
| "loss": 1.8297, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.3114754098360655e-06, | |
| "loss": 1.7981, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.344262295081967e-06, | |
| "loss": 1.8453, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 1.3770491803278687e-06, | |
| "loss": 1.7627, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.4098360655737705e-06, | |
| "loss": 1.7822, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.442622950819672e-06, | |
| "loss": 1.8156, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.4754098360655739e-06, | |
| "loss": 1.863, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.5081967213114753e-06, | |
| "loss": 1.7581, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 1.5409836065573769e-06, | |
| "loss": 1.8652, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.5737704918032787e-06, | |
| "loss": 1.8223, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.6065573770491803e-06, | |
| "loss": 1.8178, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.6393442622950819e-06, | |
| "loss": 1.8272, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.6721311475409837e-06, | |
| "loss": 1.8506, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 1.704918032786885e-06, | |
| "loss": 1.8025, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.7377049180327867e-06, | |
| "loss": 1.7876, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.7704918032786885e-06, | |
| "loss": 1.7781, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.80327868852459e-06, | |
| "loss": 1.8246, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.8360655737704917e-06, | |
| "loss": 1.7799, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.8688524590163935e-06, | |
| "loss": 1.8039, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 1.901639344262295e-06, | |
| "loss": 1.8075, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9344262295081967e-06, | |
| "loss": 1.7993, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.967213114754098e-06, | |
| "loss": 1.8053, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 2e-06, | |
| "loss": 1.7745, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9999987114893716e-06, | |
| "loss": 1.7959, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 1.9999948459608074e-06, | |
| "loss": 1.8582, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999884034242694e-06, | |
| "loss": 1.8663, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999793838963593e-06, | |
| "loss": 1.8255, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999677874003214e-06, | |
| "loss": 1.8037, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.9999536139660395e-06, | |
| "loss": 1.8094, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 1.999936863630039e-06, | |
| "loss": 1.8929, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9999175364354857e-06, | |
| "loss": 1.8412, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9998956324321866e-06, | |
| "loss": 1.7918, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.9998711516765885e-06, | |
| "loss": 1.8734, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.999844094231779e-06, | |
| "loss": 1.7952, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 1.999814460167486e-06, | |
| "loss": 1.84, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.999782249560076e-06, | |
| "loss": 1.7485, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9997474624925582e-06, | |
| "loss": 1.7845, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9997100990545777e-06, | |
| "loss": 1.8562, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9996701593424223e-06, | |
| "loss": 1.8288, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 1.9996276434590164e-06, | |
| "loss": 1.8192, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.999582551513925e-06, | |
| "loss": 1.8072, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9995348836233515e-06, | |
| "loss": 1.7816, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.999484639910136e-06, | |
| "loss": 1.7941, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.999431820503758e-06, | |
| "loss": 1.8132, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 1.9993764255403344e-06, | |
| "loss": 1.762, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.999318455162619e-06, | |
| "loss": 1.7573, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.999257909520003e-06, | |
| "loss": 1.8117, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.9991947887685134e-06, | |
| "loss": 1.7551, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.999129093070814e-06, | |
| "loss": 1.8449, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 1.999060822596204e-06, | |
| "loss": 1.7815, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.9989899775206176e-06, | |
| "loss": 1.7854, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.9989165580266245e-06, | |
| "loss": 1.7899, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.998840564303428e-06, | |
| "loss": 1.8634, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.9987619965468652e-06, | |
| "loss": 1.8095, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 1.9986808549594078e-06, | |
| "loss": 1.7522, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.9985971397501588e-06, | |
| "loss": 1.8314, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.9985108511348537e-06, | |
| "loss": 1.8314, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.998421989335861e-06, | |
| "loss": 1.7902, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.998330554582179e-06, | |
| "loss": 1.768, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 1.9982365471094366e-06, | |
| "loss": 1.8559, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9981399671598938e-06, | |
| "loss": 1.8016, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9980408149824386e-06, | |
| "loss": 1.8489, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.9979390908325886e-06, | |
| "loss": 1.8126, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.997834794972489e-06, | |
| "loss": 1.8001, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 1.997727927670912e-06, | |
| "loss": 1.8156, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.997618489203258e-06, | |
| "loss": 1.7559, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.9975064798515517e-06, | |
| "loss": 1.7617, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.997391899904443e-06, | |
| "loss": 1.8464, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.9972747496572075e-06, | |
| "loss": 1.838, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 1.997155029411744e-06, | |
| "loss": 1.7975, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.9970327394765733e-06, | |
| "loss": 1.8403, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.99690788016684e-06, | |
| "loss": 1.8571, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.9967804518043086e-06, | |
| "loss": 1.8265, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.9966504547173653e-06, | |
| "loss": 1.833, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 1.9965178892410154e-06, | |
| "loss": 1.8409, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.996382755716882e-06, | |
| "loss": 1.8173, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.9962450544932076e-06, | |
| "loss": 1.7903, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.9961047859248517e-06, | |
| "loss": 1.8099, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.9959619503732886e-06, | |
| "loss": 1.8147, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 1.995816548206609e-06, | |
| "loss": 1.7784, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9956685797995172e-06, | |
| "loss": 1.849, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9955180455333312e-06, | |
| "loss": 1.8219, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9953649457959808e-06, | |
| "loss": 1.7788, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.995209280982007e-06, | |
| "loss": 1.8205, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 1.9950510514925617e-06, | |
| "loss": 1.8305, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.9948902577354057e-06, | |
| "loss": 1.8397, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.994726900124908e-06, | |
| "loss": 1.8697, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.994560979082044e-06, | |
| "loss": 1.8091, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.994392495034396e-06, | |
| "loss": 1.8292, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 1.9942214484161517e-06, | |
| "loss": 1.8443, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.994047839668101e-06, | |
| "loss": 1.8159, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9938716692376378e-06, | |
| "loss": 1.8701, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9936929375787566e-06, | |
| "loss": 1.7886, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.9935116451520533e-06, | |
| "loss": 1.7754, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 1.993327792424722e-06, | |
| "loss": 1.812, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.993141379870555e-06, | |
| "loss": 1.8494, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9929524079699417e-06, | |
| "loss": 1.8719, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.992760877209866e-06, | |
| "loss": 1.7864, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.992566788083908e-06, | |
| "loss": 1.8107, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 1.9923701410922383e-06, | |
| "loss": 1.7902, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9921709367416207e-06, | |
| "loss": 1.7544, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.99196917554541e-06, | |
| "loss": 1.8509, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9917648580235474e-06, | |
| "loss": 1.8262, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.991557984702565e-06, | |
| "loss": 1.766, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 1.9913485561155783e-06, | |
| "loss": 1.7944, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.991136572802291e-06, | |
| "loss": 1.8283, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.990922035308987e-06, | |
| "loss": 1.7621, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9907049441885347e-06, | |
| "loss": 1.7697, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.990485300000383e-06, | |
| "loss": 1.8438, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 1.9902631033105586e-06, | |
| "loss": 1.8083, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9900383546916676e-06, | |
| "loss": 1.8385, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9898110547228923e-06, | |
| "loss": 1.7602, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.989581203989989e-06, | |
| "loss": 1.8295, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.9893488030852884e-06, | |
| "loss": 1.8248, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 1.989113852607692e-06, | |
| "loss": 1.7993, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.9888763531626725e-06, | |
| "loss": 1.8724, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.988636305362271e-06, | |
| "loss": 1.7902, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.988393709825096e-06, | |
| "loss": 1.7993, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.988148567176321e-06, | |
| "loss": 1.7969, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 1.987900878047684e-06, | |
| "loss": 1.8642, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.987650643077485e-06, | |
| "loss": 1.8064, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9873978629105853e-06, | |
| "loss": 1.7677, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9871425381984043e-06, | |
| "loss": 1.7867, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.986884669598919e-06, | |
| "loss": 1.8304, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9866242577766627e-06, | |
| "loss": 1.8228, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 1.9863613034027223e-06, | |
| "loss": 1.7832, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9860958071547365e-06, | |
| "loss": 1.8687, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9858277697168947e-06, | |
| "loss": 1.8497, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9855571917799356e-06, | |
| "loss": 1.7625, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9852840740411437e-06, | |
| "loss": 1.8641, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 1.9850084172043492e-06, | |
| "loss": 1.8542, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9847302219799266e-06, | |
| "loss": 1.8338, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9844494890847895e-06, | |
| "loss": 1.7942, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9841662192423937e-06, | |
| "loss": 1.8268, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.983880413182731e-06, | |
| "loss": 1.801, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 1.9835920716423297e-06, | |
| "loss": 1.7898, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.983301195364252e-06, | |
| "loss": 1.7926, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9830077850980933e-06, | |
| "loss": 1.8177, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9827118415999765e-06, | |
| "loss": 1.8194, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9824133656325553e-06, | |
| "loss": 1.8553, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 1.9821123579650084e-06, | |
| "loss": 1.7962, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.9818088193730385e-06, | |
| "loss": 1.8247, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.981502750638872e-06, | |
| "loss": 1.8226, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.981194152551254e-06, | |
| "loss": 1.7985, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.980883025905448e-06, | |
| "loss": 1.773, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 1.980569371503234e-06, | |
| "loss": 1.8056, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9802531901529065e-06, | |
| "loss": 1.8649, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.979934482669271e-06, | |
| "loss": 1.8389, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9796132498736443e-06, | |
| "loss": 1.7644, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.97928949259385e-06, | |
| "loss": 1.7506, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 1.9789632116642166e-06, | |
| "loss": 1.8144, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.978634407925578e-06, | |
| "loss": 1.8213, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9783030822252677e-06, | |
| "loss": 1.8077, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9779692354171195e-06, | |
| "loss": 1.7861, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9776328683614634e-06, | |
| "loss": 1.8397, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 1.9772939819251245e-06, | |
| "loss": 1.7771, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.976952576981421e-06, | |
| "loss": 1.7759, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9766086544101597e-06, | |
| "loss": 1.7946, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9762622150976374e-06, | |
| "loss": 1.8384, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9759132599366346e-06, | |
| "loss": 1.8419, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 1.9755617898264166e-06, | |
| "loss": 1.7618, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.9752078056727297e-06, | |
| "loss": 1.8551, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.974851308387798e-06, | |
| "loss": 1.7832, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.974492298890323e-06, | |
| "loss": 1.8065, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.9741307781054792e-06, | |
| "loss": 1.8013, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 1.973766746964914e-06, | |
| "loss": 1.839, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.973400206406743e-06, | |
| "loss": 1.815, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9730311573755493e-06, | |
| "loss": 1.7583, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.97265960082238e-06, | |
| "loss": 1.7778, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9722855377047442e-06, | |
| "loss": 1.811, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 1.9719089689866105e-06, | |
| "loss": 1.7852, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.9715298956384047e-06, | |
| "loss": 1.7967, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.9711483186370063e-06, | |
| "loss": 1.8403, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.970764238965748e-06, | |
| "loss": 1.7783, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.9703776576144106e-06, | |
| "loss": 1.7977, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 1.969988575579223e-06, | |
| "loss": 1.7673, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.969596993862858e-06, | |
| "loss": 1.7923, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.9692029134744295e-06, | |
| "loss": 1.844, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.9688063354294913e-06, | |
| "loss": 1.812, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.9684072607500334e-06, | |
| "loss": 1.7798, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 1.9680056904644796e-06, | |
| "loss": 1.7617, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.9676016256076855e-06, | |
| "loss": 1.7449, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.967195067220934e-06, | |
| "loss": 1.8363, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.966786016351936e-06, | |
| "loss": 1.8269, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.9663744740548232e-06, | |
| "loss": 1.7656, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 1.9659604413901496e-06, | |
| "loss": 1.8018, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.965543919424885e-06, | |
| "loss": 1.8162, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.9651249092324166e-06, | |
| "loss": 1.8419, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.9647034118925417e-06, | |
| "loss": 1.764, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.9642794284914687e-06, | |
| "loss": 1.8261, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 1.9638529601218114e-06, | |
| "loss": 1.7924, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.9634240078825876e-06, | |
| "loss": 1.7687, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.9629925728792166e-06, | |
| "loss": 1.8531, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.962558656223516e-06, | |
| "loss": 1.7934, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.9621222590336973e-06, | |
| "loss": 1.807, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 1.961683382434366e-06, | |
| "loss": 1.7776, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.961242027556517e-06, | |
| "loss": 1.8311, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9607981955375295e-06, | |
| "loss": 1.8013, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9603518875211696e-06, | |
| "loss": 1.8536, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9599031046575816e-06, | |
| "loss": 1.7611, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 1.9594518481032896e-06, | |
| "loss": 1.7873, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.95899811902119e-06, | |
| "loss": 1.7876, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.958541918580553e-06, | |
| "loss": 1.7066, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.958083247957017e-06, | |
| "loss": 1.8098, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.957622108332585e-06, | |
| "loss": 1.7821, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 1.957158500895625e-06, | |
| "loss": 1.7832, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9566924268408624e-06, | |
| "loss": 1.799, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9562238873693796e-06, | |
| "loss": 1.7274, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9557528836886135e-06, | |
| "loss": 1.8111, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.9552794170123505e-06, | |
| "loss": 1.7611, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 1.954803488560724e-06, | |
| "loss": 1.7715, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9543250995602117e-06, | |
| "loss": 1.7802, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9538442512436325e-06, | |
| "loss": 1.7697, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.953360944850143e-06, | |
| "loss": 1.7719, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9528751816252328e-06, | |
| "loss": 1.8125, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 1.9523869628207254e-06, | |
| "loss": 1.7546, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.9518962896947706e-06, | |
| "loss": 1.7955, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.951403163511843e-06, | |
| "loss": 1.8274, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.9509075855427403e-06, | |
| "loss": 1.7821, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.9504095570645768e-06, | |
| "loss": 1.8088, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 1.949909079360782e-06, | |
| "loss": 1.8297, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.9494061537210983e-06, | |
| "loss": 1.7987, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.948900781441576e-06, | |
| "loss": 1.8488, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.948392963824569e-06, | |
| "loss": 1.8249, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.9478827021787355e-06, | |
| "loss": 1.7542, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 1.9473699978190297e-06, | |
| "loss": 1.7514, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.946854852066702e-06, | |
| "loss": 1.803, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.9463372662492934e-06, | |
| "loss": 1.8222, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.9458172417006346e-06, | |
| "loss": 1.8136, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.945294779760839e-06, | |
| "loss": 1.7658, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 1.9447698817763026e-06, | |
| "loss": 1.8392, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9442425490996984e-06, | |
| "loss": 1.8257, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9437127830899744e-06, | |
| "loss": 1.7824, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.943180585112348e-06, | |
| "loss": 1.8179, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9426459565383056e-06, | |
| "loss": 1.8429, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9421088987455956e-06, | |
| "loss": 1.7787, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 1.9415694131182277e-06, | |
| "loss": 1.7805, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.9410275010464685e-06, | |
| "loss": 1.8401, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.9404831639268356e-06, | |
| "loss": 1.7954, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.9399364031620987e-06, | |
| "loss": 1.7947, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.939387220161271e-06, | |
| "loss": 1.7727, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 1.9388356163396085e-06, | |
| "loss": 1.784, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.938281593118607e-06, | |
| "loss": 1.7915, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9377251519259954e-06, | |
| "loss": 1.7829, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9371662941957347e-06, | |
| "loss": 1.8075, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.936605021368013e-06, | |
| "loss": 1.8001, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 1.9360413348892427e-06, | |
| "loss": 1.8251, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.935475236212056e-06, | |
| "loss": 1.7733, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.9349067267953e-06, | |
| "loss": 1.8052, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.934335808104037e-06, | |
| "loss": 1.7986, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.9337624816095357e-06, | |
| "loss": 1.8115, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 1.9331867487892706e-06, | |
| "loss": 1.7622, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.932608611126918e-06, | |
| "loss": 1.8088, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9320280701123507e-06, | |
| "loss": 1.7687, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.931445127241635e-06, | |
| "loss": 1.7906, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.9308597840170277e-06, | |
| "loss": 1.7157, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 1.93027204194697e-06, | |
| "loss": 1.8403, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9296819025460857e-06, | |
| "loss": 1.7787, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9290893673351777e-06, | |
| "loss": 1.728, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9284944378412205e-06, | |
| "loss": 1.7893, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9278971155973607e-06, | |
| "loss": 1.7054, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 1.9272974021429104e-06, | |
| "loss": 1.7644, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.926695299023344e-06, | |
| "loss": 1.8168, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9260908077902935e-06, | |
| "loss": 1.7614, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.925483930001546e-06, | |
| "loss": 1.7905, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9248746672210384e-06, | |
| "loss": 1.817, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 1.9242630210188545e-06, | |
| "loss": 1.8445, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9236489929712185e-06, | |
| "loss": 1.7845, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9230325846604944e-06, | |
| "loss": 1.7529, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9224137976751793e-06, | |
| "loss": 1.7691, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9217926336099004e-06, | |
| "loss": 1.7631, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 1.9211690940654107e-06, | |
| "loss": 1.7793, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.920543180648585e-06, | |
| "loss": 1.7552, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9199148949724155e-06, | |
| "loss": 1.7952, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9192842386560077e-06, | |
| "loss": 1.785, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9186512133245757e-06, | |
| "loss": 1.7388, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 1.9180158206094405e-06, | |
| "loss": 1.8321, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9173780621480214e-06, | |
| "loss": 1.7727, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9167379395838364e-06, | |
| "loss": 1.781, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.9160954545664944e-06, | |
| "loss": 1.7616, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.915450608751693e-06, | |
| "loss": 1.7842, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 1.914803403801214e-06, | |
| "loss": 1.7829, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.914153841382918e-06, | |
| "loss": 1.7974, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9135019231707414e-06, | |
| "loss": 1.7255, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.912847650844691e-06, | |
| "loss": 1.7882, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.91219102609084e-06, | |
| "loss": 1.7773, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 1.9115320506013255e-06, | |
| "loss": 1.7819, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.9108707260743404e-06, | |
| "loss": 1.7731, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.910207054214133e-06, | |
| "loss": 1.7596, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.9095410367309986e-06, | |
| "loss": 1.7473, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.908872675341279e-06, | |
| "loss": 1.7474, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 1.9082019717673562e-06, | |
| "loss": 1.7585, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.9075289277376466e-06, | |
| "loss": 1.7408, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.9068535449866002e-06, | |
| "loss": 1.7541, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.9061758252546913e-06, | |
| "loss": 1.7507, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.9054957702884189e-06, | |
| "loss": 1.8213, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 1.9048133818402991e-06, | |
| "loss": 1.7869, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9041286616688615e-06, | |
| "loss": 1.8003, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9034416115386438e-06, | |
| "loss": 1.7579, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9027522332201895e-06, | |
| "loss": 1.8254, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.902060528490041e-06, | |
| "loss": 1.7649, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 1.9013664991307362e-06, | |
| "loss": 1.8614, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.9006701469308032e-06, | |
| "loss": 1.7119, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8999714736847567e-06, | |
| "loss": 1.7665, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8992704811930923e-06, | |
| "loss": 1.7592, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.8985671712622828e-06, | |
| "loss": 1.7544, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 1.897861545704773e-06, | |
| "loss": 1.806, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8971536063389742e-06, | |
| "loss": 1.7541, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8964433549892617e-06, | |
| "loss": 1.8463, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8957307934859683e-06, | |
| "loss": 1.8462, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8950159236653805e-06, | |
| "loss": 1.7611, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 1.8942987473697322e-06, | |
| "loss": 1.7608, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8935792664472026e-06, | |
| "loss": 1.8391, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8928574827519093e-06, | |
| "loss": 1.826, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8921333981439038e-06, | |
| "loss": 1.8257, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8914070144891677e-06, | |
| "loss": 1.7911, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 1.8906783336596075e-06, | |
| "loss": 1.7824, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.8899473575330487e-06, | |
| "loss": 1.7291, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.8892140879932325e-06, | |
| "loss": 1.8055, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.8884785269298105e-06, | |
| "loss": 1.763, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.8877406762383385e-06, | |
| "loss": 1.7944, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 1.8870005378202735e-06, | |
| "loss": 1.8079, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8862581135829686e-06, | |
| "loss": 1.7776, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.885513405439666e-06, | |
| "loss": 1.8052, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.884766415309495e-06, | |
| "loss": 1.8075, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8840171451174648e-06, | |
| "loss": 1.7557, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 1.8832655967944605e-06, | |
| "loss": 1.793, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8825117722772387e-06, | |
| "loss": 1.7873, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8817556735084205e-06, | |
| "loss": 1.7842, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8809973024364888e-06, | |
| "loss": 1.7921, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.880236661015782e-06, | |
| "loss": 1.741, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 1.8794737512064888e-06, | |
| "loss": 1.7511, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8787085749746447e-06, | |
| "loss": 1.798, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8779411342921247e-06, | |
| "loss": 1.7962, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8771714311366398e-06, | |
| "loss": 1.755, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8763994674917317e-06, | |
| "loss": 1.7955, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 1.8756252453467663e-06, | |
| "loss": 1.7368, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8748487666969312e-06, | |
| "loss": 1.7901, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8740700335432287e-06, | |
| "loss": 1.8222, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8732890478924696e-06, | |
| "loss": 1.7811, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8725058117572715e-06, | |
| "loss": 1.7372, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 1.8717203271560499e-06, | |
| "loss": 1.7984, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.870932596113016e-06, | |
| "loss": 1.7976, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8701426206581693e-06, | |
| "loss": 1.7846, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8693504028272925e-06, | |
| "loss": 1.7977, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8685559446619487e-06, | |
| "loss": 1.7319, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8677592482094733e-06, | |
| "loss": 1.7326, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 1.8669603155229696e-06, | |
| "loss": 1.8116, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.8661591486613047e-06, | |
| "loss": 1.7751, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.865355749689102e-06, | |
| "loss": 1.8183, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.8645501206767383e-06, | |
| "loss": 1.7286, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.8637422637003363e-06, | |
| "loss": 1.7795, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 1.8629321808417608e-06, | |
| "loss": 1.7433, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8621198741886124e-06, | |
| "loss": 1.777, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8613053458342227e-06, | |
| "loss": 1.8075, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8604885978776488e-06, | |
| "loss": 1.7769, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8596696324236674e-06, | |
| "loss": 1.8142, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 1.8588484515827695e-06, | |
| "loss": 1.7904, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.858025057471156e-06, | |
| "loss": 1.7639, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8571994522107311e-06, | |
| "loss": 1.8408, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8563716379290965e-06, | |
| "loss": 1.7565, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8555416167595478e-06, | |
| "loss": 1.7836, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 1.8547093908410673e-06, | |
| "loss": 1.8322, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.8538749623183185e-06, | |
| "loss": 1.7878, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.8530383333416415e-06, | |
| "loss": 1.7332, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.852199506067047e-06, | |
| "loss": 1.7737, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.8513584826562107e-06, | |
| "loss": 1.7779, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 1.8505152652764679e-06, | |
| "loss": 1.8014, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8496698561008075e-06, | |
| "loss": 1.7936, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8488222573078672e-06, | |
| "loss": 1.7844, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8479724710819268e-06, | |
| "loss": 1.7462, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8471204996129037e-06, | |
| "loss": 1.7485, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 1.8462663450963464e-06, | |
| "loss": 1.8099, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8454100097334292e-06, | |
| "loss": 1.776, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8445514957309469e-06, | |
| "loss": 1.7912, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8436908053013079e-06, | |
| "loss": 1.7699, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8428279406625298e-06, | |
| "loss": 1.7629, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 1.8419629040382332e-06, | |
| "loss": 1.727, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8410956976576357e-06, | |
| "loss": 1.7597, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8402263237555469e-06, | |
| "loss": 1.748, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8393547845723616e-06, | |
| "loss": 1.8594, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.838481082354055e-06, | |
| "loss": 1.7529, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 1.8376052193521757e-06, | |
| "loss": 1.7255, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8367271978238418e-06, | |
| "loss": 1.7804, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8358470200317338e-06, | |
| "loss": 1.7526, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8349646882440876e-06, | |
| "loss": 1.7221, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.834080204734692e-06, | |
| "loss": 1.7501, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 1.8331935717828788e-06, | |
| "loss": 1.7033, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8323047916735211e-06, | |
| "loss": 1.7825, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8314138666970228e-06, | |
| "loss": 1.7489, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8305207991493178e-06, | |
| "loss": 1.7837, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.829625591331859e-06, | |
| "loss": 1.7204, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 1.8287282455516171e-06, | |
| "loss": 1.747, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.82782876412107e-06, | |
| "loss": 1.7927, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.8269271493582017e-06, | |
| "loss": 1.7864, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.8260234035864918e-06, | |
| "loss": 1.752, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.825117529134913e-06, | |
| "loss": 1.7716, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.8242095283379224e-06, | |
| "loss": 1.7956, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8232994035354574e-06, | |
| "loss": 1.8058, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8223871570729291e-06, | |
| "loss": 1.8101, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8214727913012167e-06, | |
| "loss": 1.7702, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.820556308576659e-06, | |
| "loss": 1.7393, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8196377112610524e-06, | |
| "loss": 1.7273, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.8187170017216414e-06, | |
| "loss": 1.7896, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.817794182331114e-06, | |
| "loss": 1.731, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.816869255467595e-06, | |
| "loss": 1.8067, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.8159422235146416e-06, | |
| "loss": 1.7634, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.815013088861234e-06, | |
| "loss": 1.8085, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.814081853901772e-06, | |
| "loss": 1.7552, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.8131485210360683e-06, | |
| "loss": 1.7744, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.8122130926693416e-06, | |
| "loss": 1.7522, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.81127557121221e-06, | |
| "loss": 1.7646, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.8103359590806868e-06, | |
| "loss": 1.7668, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8093942586961723e-06, | |
| "loss": 1.7521, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8084504724854484e-06, | |
| "loss": 1.7688, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8075046028806722e-06, | |
| "loss": 1.7481, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.8065566523193696e-06, | |
| "loss": 1.7742, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.80560662324443e-06, | |
| "loss": 1.784, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8046545181040972e-06, | |
| "loss": 1.771, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8037003393519679e-06, | |
| "loss": 1.7641, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.80274408944698e-06, | |
| "loss": 1.7893, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8017857708534106e-06, | |
| "loss": 1.8559, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.8008253860408663e-06, | |
| "loss": 1.7336, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.79986293748428e-06, | |
| "loss": 1.7683, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7988984276639012e-06, | |
| "loss": 1.7372, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7979318590652927e-06, | |
| "loss": 1.7963, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7969632341793225e-06, | |
| "loss": 1.8381, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7959925555021574e-06, | |
| "loss": 1.7575, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7950198255352566e-06, | |
| "loss": 1.8207, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7940450467853662e-06, | |
| "loss": 1.7789, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7930682217645119e-06, | |
| "loss": 1.7791, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.7920893529899923e-06, | |
| "loss": 1.767, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.791108442984373e-06, | |
| "loss": 1.7942, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7901254942754803e-06, | |
| "loss": 1.7598, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7891405093963937e-06, | |
| "loss": 1.8034, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.78815349088544e-06, | |
| "loss": 1.7288, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.787164441286187e-06, | |
| "loss": 1.7815, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.7861733631474364e-06, | |
| "loss": 1.797, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7851802590232182e-06, | |
| "loss": 1.8177, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7841851314727822e-06, | |
| "loss": 1.7422, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7831879830605936e-06, | |
| "loss": 1.7399, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7821888163563249e-06, | |
| "loss": 1.7812, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 1.7811876339348496e-06, | |
| "loss": 1.7869, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.780184438376237e-06, | |
| "loss": 1.7563, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7791792322657427e-06, | |
| "loss": 1.7901, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7781720181938042e-06, | |
| "loss": 1.7711, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7771627987560337e-06, | |
| "loss": 1.7734, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 1.7761515765532113e-06, | |
| "loss": 1.7324, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7751383541912782e-06, | |
| "loss": 1.7546, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7741231342813294e-06, | |
| "loss": 1.7561, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7731059194396088e-06, | |
| "loss": 1.7741, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7720867122875004e-06, | |
| "loss": 1.7245, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.7710655154515224e-06, | |
| "loss": 1.7769, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.770042331563321e-06, | |
| "loss": 1.7355, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7690171632596633e-06, | |
| "loss": 1.7452, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.767990013182429e-06, | |
| "loss": 1.7606, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7669608839786062e-06, | |
| "loss": 1.8084, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7659297783002826e-06, | |
| "loss": 1.7889, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7648966988046397e-06, | |
| "loss": 1.7204, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7638616481539448e-06, | |
| "loss": 1.7551, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7628246290155458e-06, | |
| "loss": 1.8024, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.761785644061863e-06, | |
| "loss": 1.7638, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7607446959703825e-06, | |
| "loss": 1.7361, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.75970178742365e-06, | |
| "loss": 1.7123, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7586569211092627e-06, | |
| "loss": 1.7525, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7576100997198635e-06, | |
| "loss": 1.749, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7565613259531334e-06, | |
| "loss": 1.747, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7555106025117846e-06, | |
| "loss": 1.7705, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.7544579321035538e-06, | |
| "loss": 1.7469, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7534033174411946e-06, | |
| "loss": 1.755, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7523467612424721e-06, | |
| "loss": 1.7792, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7512882662301535e-06, | |
| "loss": 1.7223, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7502278351320032e-06, | |
| "loss": 1.692, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 1.7491654706807748e-06, | |
| "loss": 1.6989, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.748101175614204e-06, | |
| "loss": 1.7227, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.7470349526750017e-06, | |
| "loss": 1.7257, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.7459668046108469e-06, | |
| "loss": 1.7522, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.7448967341743804e-06, | |
| "loss": 1.7951, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 1.02, | |
| "learning_rate": 1.743824744123196e-06, | |
| "loss": 1.6694, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.742750837219835e-06, | |
| "loss": 1.7478, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.7416750162317783e-06, | |
| "loss": 1.7157, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.7405972839314395e-06, | |
| "loss": 1.7839, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.7395176430961577e-06, | |
| "loss": 1.6986, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 1.7384360965081901e-06, | |
| "loss": 1.6926, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7373526469547055e-06, | |
| "loss": 1.7446, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7362672972277761e-06, | |
| "loss": 1.7724, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7351800501243713e-06, | |
| "loss": 1.7376, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7340909084463505e-06, | |
| "loss": 1.742, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 1.7329998750004543e-06, | |
| "loss": 1.7251, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7319069525982993e-06, | |
| "loss": 1.7639, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7308121440563696e-06, | |
| "loss": 1.7568, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7297154521960107e-06, | |
| "loss": 1.7945, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7286168798434201e-06, | |
| "loss": 1.6914, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 1.05, | |
| "learning_rate": 1.7275164298296427e-06, | |
| "loss": 1.7106, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7264141049905612e-06, | |
| "loss": 1.7053, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.72530990816689e-06, | |
| "loss": 1.7859, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7242038422041683e-06, | |
| "loss": 1.7561, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.723095909952751e-06, | |
| "loss": 1.7365, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 1.7219861142678039e-06, | |
| "loss": 1.7149, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.7208744580092928e-06, | |
| "loss": 1.7186, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.7197609440419809e-06, | |
| "loss": 1.7458, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.7186455752354165e-06, | |
| "loss": 1.7466, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.717528354463929e-06, | |
| "loss": 1.7572, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 1.07, | |
| "learning_rate": 1.7164092846066198e-06, | |
| "loss": 1.738, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.7152883685473564e-06, | |
| "loss": 1.7654, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.7141656091747623e-06, | |
| "loss": 1.6798, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.713041009382213e-06, | |
| "loss": 1.7499, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.711914572067826e-06, | |
| "loss": 1.7802, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 1.7107863001344538e-06, | |
| "loss": 1.6773, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7096561964896776e-06, | |
| "loss": 1.7358, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7085242640457986e-06, | |
| "loss": 1.7554, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7073905057198303e-06, | |
| "loss": 1.7309, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7062549244334922e-06, | |
| "loss": 1.751, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 1.7051175231132016e-06, | |
| "loss": 1.7844, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.7039783046900652e-06, | |
| "loss": 1.7343, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.7028372720998742e-06, | |
| "loss": 1.7545, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.701694428283093e-06, | |
| "loss": 1.7874, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.7005497761848546e-06, | |
| "loss": 1.7513, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 1.6994033187549524e-06, | |
| "loss": 1.7417, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.6982550589478304e-06, | |
| "loss": 1.7354, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.6971049997225793e-06, | |
| "loss": 1.7352, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.695953144042926e-06, | |
| "loss": 1.7793, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.694799494877227e-06, | |
| "loss": 1.7799, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 1.11, | |
| "learning_rate": 1.6936440551984613e-06, | |
| "loss": 1.7053, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6924868279842204e-06, | |
| "loss": 1.7682, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6913278162167044e-06, | |
| "loss": 1.7215, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6901670228827108e-06, | |
| "loss": 1.7185, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6890044509736287e-06, | |
| "loss": 1.7653, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 1.6878401034854308e-06, | |
| "loss": 1.8248, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6866739834186654e-06, | |
| "loss": 1.7018, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6855060937784484e-06, | |
| "loss": 1.6891, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6843364375744565e-06, | |
| "loss": 1.7638, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6831650178209182e-06, | |
| "loss": 1.7252, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 1.6819918375366077e-06, | |
| "loss": 1.7095, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.680816899744835e-06, | |
| "loss": 1.7227, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.6796402074734402e-06, | |
| "loss": 1.7136, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.678461763754784e-06, | |
| "loss": 1.7166, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.6772815716257411e-06, | |
| "loss": 1.7457, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 1.6760996341276914e-06, | |
| "loss": 1.7715, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6749159543065132e-06, | |
| "loss": 1.7529, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6737305352125746e-06, | |
| "loss": 1.7111, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6725433799007255e-06, | |
| "loss": 1.7536, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6713544914302908e-06, | |
| "loss": 1.7355, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 1.15, | |
| "learning_rate": 1.6701638728650611e-06, | |
| "loss": 1.7847, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6689715272732858e-06, | |
| "loss": 1.7407, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6677774577276646e-06, | |
| "loss": 1.8005, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6665816673053405e-06, | |
| "loss": 1.708, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6653841590878907e-06, | |
| "loss": 1.7602, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 1.6641849361613192e-06, | |
| "loss": 1.7986, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6629840016160493e-06, | |
| "loss": 1.7357, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6617813585469144e-06, | |
| "loss": 1.7352, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6605770100531516e-06, | |
| "loss": 1.7441, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6593709592383923e-06, | |
| "loss": 1.7138, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 1.17, | |
| "learning_rate": 1.6581632092106555e-06, | |
| "loss": 1.7629, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.6569537630823382e-06, | |
| "loss": 1.7185, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.6557426239702087e-06, | |
| "loss": 1.744, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.6545297949953993e-06, | |
| "loss": 1.6972, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.6533152792833947e-06, | |
| "loss": 1.6991, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 1.18, | |
| "learning_rate": 1.6520990799640285e-06, | |
| "loss": 1.6925, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.650881200171472e-06, | |
| "loss": 1.7568, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.6496616430442271e-06, | |
| "loss": 1.6709, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.648440411725119e-06, | |
| "loss": 1.7474, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.6472175093612862e-06, | |
| "loss": 1.7719, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.6459929391041746e-06, | |
| "loss": 1.7151, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 1.6447667041095267e-06, | |
| "loss": 1.7478, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6435388075373777e-06, | |
| "loss": 1.7257, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6423092525520421e-06, | |
| "loss": 1.699, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6410780423221095e-06, | |
| "loss": 1.7522, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6398451800204352e-06, | |
| "loss": 1.712, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 1.2, | |
| "learning_rate": 1.6386106688241309e-06, | |
| "loss": 1.7137, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.6373745119145584e-06, | |
| "loss": 1.7794, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.6361367124773207e-06, | |
| "loss": 1.7425, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.6348972737022527e-06, | |
| "loss": 1.7158, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.6336561987834151e-06, | |
| "loss": 1.7363, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 1.632413490919084e-06, | |
| "loss": 1.7048, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.631169153311744e-06, | |
| "loss": 1.6891, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6299231891680793e-06, | |
| "loss": 1.7387, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6286756016989664e-06, | |
| "loss": 1.7149, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6274263941194648e-06, | |
| "loss": 1.7447, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.6261755696488084e-06, | |
| "loss": 1.7323, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6249231315103992e-06, | |
| "loss": 1.71, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6236690829317963e-06, | |
| "loss": 1.7227, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6224134271447099e-06, | |
| "loss": 1.7161, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6211561673849913e-06, | |
| "loss": 1.6992, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 1.6198973068926259e-06, | |
| "loss": 1.7344, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.618636848911724e-06, | |
| "loss": 1.78, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.6173747966905125e-06, | |
| "loss": 1.6864, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.6161111534813267e-06, | |
| "loss": 1.7658, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.614845922540602e-06, | |
| "loss": 1.7322, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 1.24, | |
| "learning_rate": 1.6135791071288657e-06, | |
| "loss": 1.7848, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.612310710510728e-06, | |
| "loss": 1.7114, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6110407359548733e-06, | |
| "loss": 1.7484, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6097691867340543e-06, | |
| "loss": 1.7689, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6084960661250793e-06, | |
| "loss": 1.7018, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 1.6072213774088073e-06, | |
| "loss": 1.7428, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.6059451238701388e-06, | |
| "loss": 1.7393, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.6046673087980055e-06, | |
| "loss": 1.7786, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.6033879354853648e-06, | |
| "loss": 1.7473, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.602107007229188e-06, | |
| "loss": 1.7678, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 1.26, | |
| "learning_rate": 1.6008245273304554e-06, | |
| "loss": 1.727, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5995404990941445e-06, | |
| "loss": 1.7662, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5982549258292233e-06, | |
| "loss": 1.7225, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5969678108486417e-06, | |
| "loss": 1.7646, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.595679157469322e-06, | |
| "loss": 1.6826, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 1.5943889690121518e-06, | |
| "loss": 1.7319, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.5930972488019736e-06, | |
| "loss": 1.737, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.5918040001675783e-06, | |
| "loss": 1.7317, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.5905092264416951e-06, | |
| "loss": 1.8041, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.5892129309609829e-06, | |
| "loss": 1.7459, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 1.5879151170660236e-06, | |
| "loss": 1.688, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.5866157881013105e-06, | |
| "loss": 1.712, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.585314947415242e-06, | |
| "loss": 1.696, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.5840125983601124e-06, | |
| "loss": 1.7379, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.582708744292103e-06, | |
| "loss": 1.712, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 1.5814033885712731e-06, | |
| "loss": 1.6724, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.5800965345615524e-06, | |
| "loss": 1.7081, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.578788185630732e-06, | |
| "loss": 1.7365, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.5774783451504536e-06, | |
| "loss": 1.7326, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.576167016496205e-06, | |
| "loss": 1.7431, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 1.3, | |
| "learning_rate": 1.5748542030473076e-06, | |
| "loss": 1.7429, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.5741972405819502e-06, | |
| "loss": 1.7743, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.5728822062856757e-06, | |
| "loss": 1.747, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.5715656956597782e-06, | |
| "loss": 1.7156, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.570247712096934e-06, | |
| "loss": 1.7401, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 1.568928258993614e-06, | |
| "loss": 1.7241, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5676073397500774e-06, | |
| "loss": 1.741, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.566284957770361e-06, | |
| "loss": 1.7617, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5649611164622712e-06, | |
| "loss": 1.7222, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5636358192373752e-06, | |
| "loss": 1.758, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 1.5623090695109923e-06, | |
| "loss": 1.7101, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5609808707021848e-06, | |
| "loss": 1.7094, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5596512262337489e-06, | |
| "loss": 1.7366, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5583201395322068e-06, | |
| "loss": 1.7033, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5569876140277976e-06, | |
| "loss": 1.7713, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 1.33, | |
| "learning_rate": 1.5556536531544674e-06, | |
| "loss": 1.7216, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.5543182603498615e-06, | |
| "loss": 1.6945, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.5529814390553162e-06, | |
| "loss": 1.7828, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.551643192715848e-06, | |
| "loss": 1.712, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.5503035247801464e-06, | |
| "loss": 1.7641, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 1.5489624387005638e-06, | |
| "loss": 1.7353, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5476199379331078e-06, | |
| "loss": 1.7454, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5462760259374314e-06, | |
| "loss": 1.7672, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5449307061768245e-06, | |
| "loss": 1.7404, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5435839821182043e-06, | |
| "loss": 1.749, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 1.35, | |
| "learning_rate": 1.5422358572321078e-06, | |
| "loss": 1.7655, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.540886334992681e-06, | |
| "loss": 1.6808, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.5395354188776718e-06, | |
| "loss": 1.7245, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.5381831123684194e-06, | |
| "loss": 1.7344, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.5368294189498471e-06, | |
| "loss": 1.7266, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 1.5354743421104508e-06, | |
| "loss": 1.7561, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5341178853422928e-06, | |
| "loss": 1.7185, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5327600521409906e-06, | |
| "loss": 1.6742, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5314008460057098e-06, | |
| "loss": 1.7401, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5300402704391533e-06, | |
| "loss": 1.7582, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 1.5286783289475527e-06, | |
| "loss": 1.7308, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.527315025040661e-06, | |
| "loss": 1.6841, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.5259503622317408e-06, | |
| "loss": 1.7609, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.5245843440375574e-06, | |
| "loss": 1.7415, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.523216973978369e-06, | |
| "loss": 1.7465, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 1.5218482555779164e-06, | |
| "loss": 1.7168, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.520478192363417e-06, | |
| "loss": 1.7005, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.5191067878655522e-06, | |
| "loss": 1.7353, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.517734045618461e-06, | |
| "loss": 1.7134, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.5163599691597288e-06, | |
| "loss": 1.7103, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 1.39, | |
| "learning_rate": 1.5149845620303801e-06, | |
| "loss": 1.7121, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.513607827774869e-06, | |
| "loss": 1.7294, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5122297699410678e-06, | |
| "loss": 1.757, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5108503920802613e-06, | |
| "loss": 1.7124, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5094696977471358e-06, | |
| "loss": 1.6962, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 1.5080876904997694e-06, | |
| "loss": 1.7188, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.506704373899625e-06, | |
| "loss": 1.7219, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.505319751511538e-06, | |
| "loss": 1.7692, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.5039338269037103e-06, | |
| "loss": 1.6748, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.5025466036476985e-06, | |
| "loss": 1.7626, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.5011580853184073e-06, | |
| "loss": 1.7258, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 1.41, | |
| "learning_rate": 1.4997682754940773e-06, | |
| "loss": 1.7283, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.498377177756278e-06, | |
| "loss": 1.7183, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.496984795689898e-06, | |
| "loss": 1.739, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.4955911328831353e-06, | |
| "loss": 1.7227, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.4941961929274892e-06, | |
| "loss": 1.6902, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 1.4927999794177486e-06, | |
| "loss": 1.773, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4914024959519861e-06, | |
| "loss": 1.7376, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.490003746131546e-06, | |
| "loss": 1.7352, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4886037335610366e-06, | |
| "loss": 1.7633, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4872024618483199e-06, | |
| "loss": 1.7321, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 1.4857999346045026e-06, | |
| "loss": 1.7696, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.484396155443928e-06, | |
| "loss": 1.7155, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.482991127984164e-06, | |
| "loss": 1.686, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.4815848558459965e-06, | |
| "loss": 1.6871, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.4801773426534188e-06, | |
| "loss": 1.7717, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 1.44, | |
| "learning_rate": 1.478768592033622e-06, | |
| "loss": 1.7399, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.4773586076169872e-06, | |
| "loss": 1.7117, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.4759473930370736e-06, | |
| "loss": 1.7231, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.474534951930611e-06, | |
| "loss": 1.7653, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.4731212879374905e-06, | |
| "loss": 1.7456, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 1.471706404700754e-06, | |
| "loss": 1.7219, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.470290305866586e-06, | |
| "loss": 1.6839, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4688729950843033e-06, | |
| "loss": 1.7499, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4674544760063455e-06, | |
| "loss": 1.7668, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.4660347522882668e-06, | |
| "loss": 1.7368, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 1.464613827588725e-06, | |
| "loss": 1.7342, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4631917055694735e-06, | |
| "loss": 1.7864, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4617683898953514e-06, | |
| "loss": 1.7277, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4603438842342728e-06, | |
| "loss": 1.7806, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4589181922572194e-06, | |
| "loss": 1.7868, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 1.4574913176382296e-06, | |
| "loss": 1.7212, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.4560632640543893e-06, | |
| "loss": 1.7262, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.4546340351858237e-06, | |
| "loss": 1.7261, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.4532036347156854e-06, | |
| "loss": 1.7392, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.451772066330147e-06, | |
| "loss": 1.6957, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 1.48, | |
| "learning_rate": 1.4503393337183906e-06, | |
| "loss": 1.6911, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.448905440572599e-06, | |
| "loss": 1.6689, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4474703905879447e-06, | |
| "loss": 1.7359, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.446034187462582e-06, | |
| "loss": 1.7313, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4445968348976373e-06, | |
| "loss": 1.8013, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 1.4431583365971992e-06, | |
| "loss": 1.7201, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.441718696268307e-06, | |
| "loss": 1.742, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.4402779176209457e-06, | |
| "loss": 1.7234, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.4388360043680314e-06, | |
| "loss": 1.6799, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.4373929602254061e-06, | |
| "loss": 1.7464, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 1.5, | |
| "learning_rate": 1.435948788911825e-06, | |
| "loss": 1.777, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.4345034941489482e-06, | |
| "loss": 1.672, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.433057079661331e-06, | |
| "loss": 1.7038, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.4316095491764145e-06, | |
| "loss": 1.7479, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.430160906424515e-06, | |
| "loss": 1.7366, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 1.428711155138816e-06, | |
| "loss": 1.7715, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4272602990553574e-06, | |
| "loss": 1.7368, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4258083419130264e-06, | |
| "loss": 1.6916, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.424355287453547e-06, | |
| "loss": 1.7251, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4229011394214717e-06, | |
| "loss": 1.7461, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 1.4214459015641708e-06, | |
| "loss": 1.7513, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.4199895776318228e-06, | |
| "loss": 1.772, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.4185321713774059e-06, | |
| "loss": 1.6761, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.4170736865566872e-06, | |
| "loss": 1.7396, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.4156141269282124e-06, | |
| "loss": 1.7004, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 1.4141534962532984e-06, | |
| "loss": 1.7606, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4126917982960212e-06, | |
| "loss": 1.7303, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4112290368232072e-06, | |
| "loss": 1.7514, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4097652156044242e-06, | |
| "loss": 1.7644, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4083003384119704e-06, | |
| "loss": 1.7305, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 1.54, | |
| "learning_rate": 1.4068344090208656e-06, | |
| "loss": 1.6838, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.405367431208841e-06, | |
| "loss": 1.7386, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.4038994087563297e-06, | |
| "loss": 1.7237, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.4024303454464564e-06, | |
| "loss": 1.7457, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.4009602450650287e-06, | |
| "loss": 1.7415, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 1.3994891114005264e-06, | |
| "loss": 1.7358, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.398016948244093e-06, | |
| "loss": 1.7364, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.3965437593895226e-06, | |
| "loss": 1.7275, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.395069548633256e-06, | |
| "loss": 1.7275, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.3935943197743642e-06, | |
| "loss": 1.7671, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 1.56, | |
| "learning_rate": 1.3921180766145443e-06, | |
| "loss": 1.7179, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3906408229581062e-06, | |
| "loss": 1.6914, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3891625626119632e-06, | |
| "loss": 1.7091, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3876832993856246e-06, | |
| "loss": 1.6798, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3862030370911827e-06, | |
| "loss": 1.7547, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 1.3847217795433047e-06, | |
| "loss": 1.7722, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.3832395305592232e-06, | |
| "loss": 1.7308, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.381756293958725e-06, | |
| "loss": 1.6963, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.3802720735641432e-06, | |
| "loss": 1.794, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.378786873200344e-06, | |
| "loss": 1.6798, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 1.377300696694721e-06, | |
| "loss": 1.7114, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.3758135478771828e-06, | |
| "loss": 1.7843, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.374325430580143e-06, | |
| "loss": 1.7503, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.3728363486385122e-06, | |
| "loss": 1.7296, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.3713463058896856e-06, | |
| "loss": 1.7322, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 1.59, | |
| "learning_rate": 1.3698553061735351e-06, | |
| "loss": 1.696, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.368363353332399e-06, | |
| "loss": 1.791, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.3668704512110714e-06, | |
| "loss": 1.7301, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.3653766036567923e-06, | |
| "loss": 1.6983, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.3638818145192395e-06, | |
| "loss": 1.7793, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 1.3623860876505155e-06, | |
| "loss": 1.735, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.360889426905141e-06, | |
| "loss": 1.7408, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.359391836140042e-06, | |
| "loss": 1.7122, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.357893319214542e-06, | |
| "loss": 1.6859, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.3563938799903505e-06, | |
| "loss": 1.7284, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 1.354893522331555e-06, | |
| "loss": 1.7578, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.3533922501046087e-06, | |
| "loss": 1.7275, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.351890067178322e-06, | |
| "loss": 1.7473, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.350386977423852e-06, | |
| "loss": 1.7252, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.3488829847146936e-06, | |
| "loss": 1.7676, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.3473780929266675e-06, | |
| "loss": 1.6839, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 1.345872305937912e-06, | |
| "loss": 1.755, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3443656276288723e-06, | |
| "loss": 1.7246, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3428580618822897e-06, | |
| "loss": 1.7662, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3413496125831944e-06, | |
| "loss": 1.7334, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3398402836188912e-06, | |
| "loss": 1.7078, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 1.63, | |
| "learning_rate": 1.3383300788789534e-06, | |
| "loss": 1.7083, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.3368190022552108e-06, | |
| "loss": 1.751, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.3353070576417395e-06, | |
| "loss": 1.7763, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.3337942489348537e-06, | |
| "loss": 1.7089, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.332280580033093e-06, | |
| "loss": 1.7138, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 1.330766054837214e-06, | |
| "loss": 1.7001, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.3292506772501816e-06, | |
| "loss": 1.7434, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.327734451177155e-06, | |
| "loss": 1.7676, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.326217380525481e-06, | |
| "loss": 1.7203, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.3246994692046835e-06, | |
| "loss": 1.7887, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 1.65, | |
| "learning_rate": 1.3231807211264518e-06, | |
| "loss": 1.705, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.321661140204632e-06, | |
| "loss": 1.7378, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.320140730355217e-06, | |
| "loss": 1.6944, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.3186194954963345e-06, | |
| "loss": 1.6968, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.3170974395482395e-06, | |
| "loss": 1.7132, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 1.3155745664333023e-06, | |
| "loss": 1.7319, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3140508800759996e-06, | |
| "loss": 1.7053, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3125263844029032e-06, | |
| "loss": 1.7298, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3110010833426712e-06, | |
| "loss": 1.728, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3094749808260365e-06, | |
| "loss": 1.7833, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 1.3079480807857983e-06, | |
| "loss": 1.731, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.3064203871568098e-06, | |
| "loss": 1.7514, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.3048919038759705e-06, | |
| "loss": 1.7312, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.3033626348822135e-06, | |
| "loss": 1.716, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.3018325841164984e-06, | |
| "loss": 1.7134, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 1.300301755521798e-06, | |
| "loss": 1.761, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.2987701530430903e-06, | |
| "loss": 1.7453, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.2972377806273478e-06, | |
| "loss": 1.7532, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.295704642223526e-06, | |
| "loss": 1.6817, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.2941707417825556e-06, | |
| "loss": 1.7187, | |
| "step": 1708 | |
| }, | |
| { | |
| "epoch": 1.69, | |
| "learning_rate": 1.292636083257331e-06, | |
| "loss": 1.705, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.291100670602699e-06, | |
| "loss": 1.7183, | |
| "step": 1712 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.2895645077754515e-06, | |
| "loss": 1.6986, | |
| "step": 1714 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.288027598734312e-06, | |
| "loss": 1.7339, | |
| "step": 1716 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.2864899474399287e-06, | |
| "loss": 1.7451, | |
| "step": 1718 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 1.284951557854861e-06, | |
| "loss": 1.7132, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.2834124339435712e-06, | |
| "loss": 1.6633, | |
| "step": 1722 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.2818725796724153e-06, | |
| "loss": 1.6976, | |
| "step": 1724 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.2803319990096299e-06, | |
| "loss": 1.729, | |
| "step": 1726 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.2787906959253242e-06, | |
| "loss": 1.7273, | |
| "step": 1728 | |
| }, | |
| { | |
| "epoch": 1.71, | |
| "learning_rate": 1.277248674391469e-06, | |
| "loss": 1.7227, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.2757059383818861e-06, | |
| "loss": 1.7268, | |
| "step": 1732 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.2741624918722402e-06, | |
| "loss": 1.6966, | |
| "step": 1734 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.272618338840025e-06, | |
| "loss": 1.7406, | |
| "step": 1736 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.2710734832645555e-06, | |
| "loss": 1.7516, | |
| "step": 1738 | |
| }, | |
| { | |
| "epoch": 1.72, | |
| "learning_rate": 1.2695279291269576e-06, | |
| "loss": 1.7037, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.2679816804101573e-06, | |
| "loss": 1.7144, | |
| "step": 1742 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.2664347410988701e-06, | |
| "loss": 1.7097, | |
| "step": 1744 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.2648871151795919e-06, | |
| "loss": 1.7297, | |
| "step": 1746 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.2633388066405873e-06, | |
| "loss": 1.768, | |
| "step": 1748 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 1.26178981947188e-06, | |
| "loss": 1.7564, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2602401576652435e-06, | |
| "loss": 1.7013, | |
| "step": 1752 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2586898252141892e-06, | |
| "loss": 1.7277, | |
| "step": 1754 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2571388261139562e-06, | |
| "loss": 1.6789, | |
| "step": 1756 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2555871643615024e-06, | |
| "loss": 1.6935, | |
| "step": 1758 | |
| }, | |
| { | |
| "epoch": 1.74, | |
| "learning_rate": 1.2540348439554932e-06, | |
| "loss": 1.7629, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2524818688962917e-06, | |
| "loss": 1.7519, | |
| "step": 1762 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2509282431859468e-06, | |
| "loss": 1.7405, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2493739708281857e-06, | |
| "loss": 1.6889, | |
| "step": 1766 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2478190558284007e-06, | |
| "loss": 1.6996, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 1.2462635021936412e-06, | |
| "loss": 1.6835, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.244707313932602e-06, | |
| "loss": 1.721, | |
| "step": 1772 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.243150495055613e-06, | |
| "loss": 1.7609, | |
| "step": 1774 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.24159304957463e-06, | |
| "loss": 1.685, | |
| "step": 1776 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.2400349815032226e-06, | |
| "loss": 1.7116, | |
| "step": 1778 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 1.2384762948565657e-06, | |
| "loss": 1.7281, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2369169936514281e-06, | |
| "loss": 1.7215, | |
| "step": 1782 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2353570819061615e-06, | |
| "loss": 1.6872, | |
| "step": 1784 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2337965636406922e-06, | |
| "loss": 1.6888, | |
| "step": 1786 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2322354428765087e-06, | |
| "loss": 1.7483, | |
| "step": 1788 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 1.2306737236366524e-06, | |
| "loss": 1.6962, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.229111409945707e-06, | |
| "loss": 1.727, | |
| "step": 1792 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.2275485058297878e-06, | |
| "loss": 1.7225, | |
| "step": 1794 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.225985015316533e-06, | |
| "loss": 1.7546, | |
| "step": 1796 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.2244209424350895e-06, | |
| "loss": 1.7603, | |
| "step": 1798 | |
| }, | |
| { | |
| "epoch": 1.78, | |
| "learning_rate": 1.222856291216107e-06, | |
| "loss": 1.7435, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.2212910656917252e-06, | |
| "loss": 1.6999, | |
| "step": 1802 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.2197252698955629e-06, | |
| "loss": 1.7197, | |
| "step": 1804 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.21815890786271e-06, | |
| "loss": 1.7252, | |
| "step": 1806 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.2165919836297141e-06, | |
| "loss": 1.7659, | |
| "step": 1808 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 1.215024501234572e-06, | |
| "loss": 1.7596, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2134564647167203e-06, | |
| "loss": 1.7666, | |
| "step": 1812 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2118878781170213e-06, | |
| "loss": 1.7682, | |
| "step": 1814 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2103187454777564e-06, | |
| "loss": 1.7111, | |
| "step": 1816 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.208749070842614e-06, | |
| "loss": 1.7394, | |
| "step": 1818 | |
| }, | |
| { | |
| "epoch": 1.8, | |
| "learning_rate": 1.2071788582566788e-06, | |
| "loss": 1.7801, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.2056081117664217e-06, | |
| "loss": 1.7963, | |
| "step": 1822 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.2040368354196904e-06, | |
| "loss": 1.7514, | |
| "step": 1824 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.202465033265697e-06, | |
| "loss": 1.7433, | |
| "step": 1826 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.2008927093550092e-06, | |
| "loss": 1.7135, | |
| "step": 1828 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 1.1993198677395392e-06, | |
| "loss": 1.7861, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1977465124725331e-06, | |
| "loss": 1.6881, | |
| "step": 1832 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1961726476085617e-06, | |
| "loss": 1.7643, | |
| "step": 1834 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1945982772035066e-06, | |
| "loss": 1.7258, | |
| "step": 1836 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1930234053145551e-06, | |
| "loss": 1.7356, | |
| "step": 1838 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 1.1914480360001848e-06, | |
| "loss": 1.7001, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.1898721733201566e-06, | |
| "loss": 1.6947, | |
| "step": 1842 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.1882958213355018e-06, | |
| "loss": 1.7146, | |
| "step": 1844 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.1867189841085128e-06, | |
| "loss": 1.7223, | |
| "step": 1846 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.185141665702733e-06, | |
| "loss": 1.7309, | |
| "step": 1848 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 1.1835638701829447e-06, | |
| "loss": 1.7528, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1819856016151615e-06, | |
| "loss": 1.746, | |
| "step": 1852 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1804068640666146e-06, | |
| "loss": 1.731, | |
| "step": 1854 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1788276616057437e-06, | |
| "loss": 1.7242, | |
| "step": 1856 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1772479983021883e-06, | |
| "loss": 1.6927, | |
| "step": 1858 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1756678782267736e-06, | |
| "loss": 1.7216, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 1.84, | |
| "learning_rate": 1.1740873054515021e-06, | |
| "loss": 1.724, | |
| "step": 1862 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1725062840495444e-06, | |
| "loss": 1.7287, | |
| "step": 1864 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1709248180952257e-06, | |
| "loss": 1.755, | |
| "step": 1866 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1693429116640176e-06, | |
| "loss": 1.7372, | |
| "step": 1868 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1677605688325264e-06, | |
| "loss": 1.7581, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 1.1661777936784835e-06, | |
| "loss": 1.7146, | |
| "step": 1872 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.164594590280734e-06, | |
| "loss": 1.7697, | |
| "step": 1874 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.1630109627192263e-06, | |
| "loss": 1.7321, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.1614269150750032e-06, | |
| "loss": 1.7004, | |
| "step": 1878 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.1598424514301884e-06, | |
| "loss": 1.7181, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 1.1582575758679785e-06, | |
| "loss": 1.7547, | |
| "step": 1882 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.1566722924726316e-06, | |
| "loss": 1.7581, | |
| "step": 1884 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.1550866053294566e-06, | |
| "loss": 1.7138, | |
| "step": 1886 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.1535005185248032e-06, | |
| "loss": 1.7351, | |
| "step": 1888 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.151914036146051e-06, | |
| "loss": 1.6995, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 1.87, | |
| "learning_rate": 1.1503271622815983e-06, | |
| "loss": 1.7171, | |
| "step": 1892 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1487399010208528e-06, | |
| "loss": 1.6896, | |
| "step": 1894 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.147152256454221e-06, | |
| "loss": 1.6964, | |
| "step": 1896 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1455642326730962e-06, | |
| "loss": 1.6868, | |
| "step": 1898 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1439758337698497e-06, | |
| "loss": 1.753, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 1.1423870638378192e-06, | |
| "loss": 1.7063, | |
| "step": 1902 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1407979269712984e-06, | |
| "loss": 1.7445, | |
| "step": 1904 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1392084272655273e-06, | |
| "loss": 1.6722, | |
| "step": 1906 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1376185688166797e-06, | |
| "loss": 1.7483, | |
| "step": 1908 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1360283557218555e-06, | |
| "loss": 1.7225, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 1.1344377920790665e-06, | |
| "loss": 1.7167, | |
| "step": 1912 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.13284688198723e-06, | |
| "loss": 1.7582, | |
| "step": 1914 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.1312556295461547e-06, | |
| "loss": 1.7406, | |
| "step": 1916 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.129664038856532e-06, | |
| "loss": 1.7747, | |
| "step": 1918 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.1280721140199255e-06, | |
| "loss": 1.6814, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 1.126479859138758e-06, | |
| "loss": 1.7328, | |
| "step": 1922 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.1248872783163052e-06, | |
| "loss": 1.6897, | |
| "step": 1924 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.1232943756566821e-06, | |
| "loss": 1.6813, | |
| "step": 1926 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.1217011552648315e-06, | |
| "loss": 1.6974, | |
| "step": 1928 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.1201076212465174e-06, | |
| "loss": 1.6851, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 1.91, | |
| "learning_rate": 1.11851377770831e-06, | |
| "loss": 1.726, | |
| "step": 1932 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.1169196287575783e-06, | |
| "loss": 1.7227, | |
| "step": 1934 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.115325178502478e-06, | |
| "loss": 1.7094, | |
| "step": 1936 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.1137304310519413e-06, | |
| "loss": 1.7229, | |
| "step": 1938 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.1121353905156662e-06, | |
| "loss": 1.729, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 1.1105400610041064e-06, | |
| "loss": 1.6949, | |
| "step": 1942 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1089444466284597e-06, | |
| "loss": 1.7241, | |
| "step": 1944 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1073485515006579e-06, | |
| "loss": 1.6917, | |
| "step": 1946 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1057523797333575e-06, | |
| "loss": 1.7039, | |
| "step": 1948 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1041559354399263e-06, | |
| "loss": 1.7494, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 1.93, | |
| "learning_rate": 1.1025592227344355e-06, | |
| "loss": 1.6907, | |
| "step": 1952 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.1009622457316479e-06, | |
| "loss": 1.7057, | |
| "step": 1954 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.099365008547007e-06, | |
| "loss": 1.6984, | |
| "step": 1956 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.097767515296627e-06, | |
| "loss": 1.6396, | |
| "step": 1958 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.0961697700972813e-06, | |
| "loss": 1.684, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 1.0945717770663943e-06, | |
| "loss": 1.7735, | |
| "step": 1962 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.0929735403220274e-06, | |
| "loss": 1.7183, | |
| "step": 1964 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.0913750639828709e-06, | |
| "loss": 1.7382, | |
| "step": 1966 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.0897763521682325e-06, | |
| "loss": 1.7274, | |
| "step": 1968 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.0881774089980257e-06, | |
| "loss": 1.7601, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 1.086578238592762e-06, | |
| "loss": 1.7263, | |
| "step": 1972 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0849788450735369e-06, | |
| "loss": 1.7362, | |
| "step": 1974 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0833792325620217e-06, | |
| "loss": 1.7737, | |
| "step": 1976 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0817794051804522e-06, | |
| "loss": 1.7052, | |
| "step": 1978 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0801793670516166e-06, | |
| "loss": 1.7665, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 1.0785791222988485e-06, | |
| "loss": 1.7011, | |
| "step": 1982 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0769786750460115e-06, | |
| "loss": 1.7428, | |
| "step": 1984 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0753780294174925e-06, | |
| "loss": 1.6434, | |
| "step": 1986 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0737771895381896e-06, | |
| "loss": 1.7, | |
| "step": 1988 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.072176159533501e-06, | |
| "loss": 1.6991, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 1.97, | |
| "learning_rate": 1.0705749435293151e-06, | |
| "loss": 1.7181, | |
| "step": 1992 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0689735456519997e-06, | |
| "loss": 1.7335, | |
| "step": 1994 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0673719700283908e-06, | |
| "loss": 1.7361, | |
| "step": 1996 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0657702207857834e-06, | |
| "loss": 1.7084, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0641683020519188e-06, | |
| "loss": 1.7251, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 1.0625662179549756e-06, | |
| "loss": 1.7079, | |
| "step": 2002 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.060963972623559e-06, | |
| "loss": 1.7479, | |
| "step": 2004 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.0593615701866891e-06, | |
| "loss": 1.6626, | |
| "step": 2006 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.057759014773791e-06, | |
| "loss": 1.6945, | |
| "step": 2008 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.056156310514684e-06, | |
| "loss": 1.7186, | |
| "step": 2010 | |
| }, | |
| { | |
| "epoch": 1.99, | |
| "learning_rate": 1.0545534615395713e-06, | |
| "loss": 1.6669, | |
| "step": 2012 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0529504719790286e-06, | |
| "loss": 1.7381, | |
| "step": 2014 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.051347345963994e-06, | |
| "loss": 1.6247, | |
| "step": 2016 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0497440876257573e-06, | |
| "loss": 1.737, | |
| "step": 2018 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0481407010959493e-06, | |
| "loss": 1.7018, | |
| "step": 2020 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "learning_rate": 1.0465371905065315e-06, | |
| "loss": 1.7111, | |
| "step": 2022 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0449335599897847e-06, | |
| "loss": 1.719, | |
| "step": 2024 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0433298136782983e-06, | |
| "loss": 1.7134, | |
| "step": 2026 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0417259557049613e-06, | |
| "loss": 1.7313, | |
| "step": 2028 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0401219902029497e-06, | |
| "loss": 1.7178, | |
| "step": 2030 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 1.0385179213057158e-06, | |
| "loss": 1.6614, | |
| "step": 2032 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0369137531469805e-06, | |
| "loss": 1.7438, | |
| "step": 2034 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0353094898607183e-06, | |
| "loss": 1.6895, | |
| "step": 2036 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0337051355811506e-06, | |
| "loss": 1.7036, | |
| "step": 2038 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0321006944427323e-06, | |
| "loss": 1.6317, | |
| "step": 2040 | |
| }, | |
| { | |
| "epoch": 2.02, | |
| "learning_rate": 1.0304961705801413e-06, | |
| "loss": 1.7419, | |
| "step": 2042 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0288915681282708e-06, | |
| "loss": 1.6857, | |
| "step": 2044 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0272868912222154e-06, | |
| "loss": 1.6882, | |
| "step": 2046 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.025682143997261e-06, | |
| "loss": 1.6892, | |
| "step": 2048 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0240773305888757e-06, | |
| "loss": 1.733, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 1.0224724551326975e-06, | |
| "loss": 1.6998, | |
| "step": 2052 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.020867521764525e-06, | |
| "loss": 1.725, | |
| "step": 2054 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.0192625346203053e-06, | |
| "loss": 1.6571, | |
| "step": 2056 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.0176574978361241e-06, | |
| "loss": 1.7124, | |
| "step": 2058 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.0160524155481963e-06, | |
| "loss": 1.6705, | |
| "step": 2060 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 1.014447291892852e-06, | |
| "loss": 1.7278, | |
| "step": 2062 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0128421310065295e-06, | |
| "loss": 1.6724, | |
| "step": 2064 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.011236937025762e-06, | |
| "loss": 1.6778, | |
| "step": 2066 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0096317140871693e-06, | |
| "loss": 1.7119, | |
| "step": 2068 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0080264663274447e-06, | |
| "loss": 1.7009, | |
| "step": 2070 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0064211978833456e-06, | |
| "loss": 1.7222, | |
| "step": 2072 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 1.0048159128916832e-06, | |
| "loss": 1.6838, | |
| "step": 2074 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.0032106154893109e-06, | |
| "loss": 1.6765, | |
| "step": 2076 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1.0016053098131143e-06, | |
| "loss": 1.6718, | |
| "step": 2078 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 1e-06, | |
| "loss": 1.6745, | |
| "step": 2080 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 9.983946901868858e-07, | |
| "loss": 1.6932, | |
| "step": 2082 | |
| }, | |
| { | |
| "epoch": 2.06, | |
| "learning_rate": 9.967893845106888e-07, | |
| "loss": 1.6402, | |
| "step": 2084 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.95184087108317e-07, | |
| "loss": 1.7658, | |
| "step": 2086 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.935788021166543e-07, | |
| "loss": 1.7063, | |
| "step": 2088 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.919735336725554e-07, | |
| "loss": 1.6647, | |
| "step": 2090 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.903682859128306e-07, | |
| "loss": 1.7114, | |
| "step": 2092 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 9.887630629742376e-07, | |
| "loss": 1.7583, | |
| "step": 2094 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.87157868993471e-07, | |
| "loss": 1.7003, | |
| "step": 2096 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.85552708107148e-07, | |
| "loss": 1.6645, | |
| "step": 2098 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.839475844518038e-07, | |
| "loss": 1.7179, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.823425021638758e-07, | |
| "loss": 1.7364, | |
| "step": 2102 | |
| }, | |
| { | |
| "epoch": 2.08, | |
| "learning_rate": 9.807374653796948e-07, | |
| "loss": 1.688, | |
| "step": 2104 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.79132478235475e-07, | |
| "loss": 1.6513, | |
| "step": 2106 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.775275448673026e-07, | |
| "loss": 1.6897, | |
| "step": 2108 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.759226694111242e-07, | |
| "loss": 1.7039, | |
| "step": 2110 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.74317856002739e-07, | |
| "loss": 1.6901, | |
| "step": 2112 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 9.727131087777848e-07, | |
| "loss": 1.7266, | |
| "step": 2114 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.711084318717291e-07, | |
| "loss": 1.6732, | |
| "step": 2116 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.695038294198588e-07, | |
| "loss": 1.7003, | |
| "step": 2118 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.678993055572678e-07, | |
| "loss": 1.7708, | |
| "step": 2120 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.66294864418849e-07, | |
| "loss": 1.6985, | |
| "step": 2122 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 9.646905101392816e-07, | |
| "loss": 1.6716, | |
| "step": 2124 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.630862468530196e-07, | |
| "loss": 1.6845, | |
| "step": 2126 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.614820786942841e-07, | |
| "loss": 1.7186, | |
| "step": 2128 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.598780097970505e-07, | |
| "loss": 1.7135, | |
| "step": 2130 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.582740442950386e-07, | |
| "loss": 1.7215, | |
| "step": 2132 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 9.566701863217018e-07, | |
| "loss": 1.676, | |
| "step": 2134 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.550664400102154e-07, | |
| "loss": 1.7331, | |
| "step": 2136 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.534628094934685e-07, | |
| "loss": 1.7449, | |
| "step": 2138 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.518592989040506e-07, | |
| "loss": 1.6794, | |
| "step": 2140 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.502559123742426e-07, | |
| "loss": 1.6426, | |
| "step": 2142 | |
| }, | |
| { | |
| "epoch": 2.12, | |
| "learning_rate": 9.486526540360062e-07, | |
| "loss": 1.7009, | |
| "step": 2144 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.470495280209716e-07, | |
| "loss": 1.6821, | |
| "step": 2146 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.454465384604287e-07, | |
| "loss": 1.7464, | |
| "step": 2148 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.43843689485316e-07, | |
| "loss": 1.701, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.422409852262092e-07, | |
| "loss": 1.7189, | |
| "step": 2152 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 9.406384298133111e-07, | |
| "loss": 1.7062, | |
| "step": 2154 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.39036027376441e-07, | |
| "loss": 1.6925, | |
| "step": 2156 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.374337820450243e-07, | |
| "loss": 1.6618, | |
| "step": 2158 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.358316979480813e-07, | |
| "loss": 1.6962, | |
| "step": 2160 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.34229779214217e-07, | |
| "loss": 1.6879, | |
| "step": 2162 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 9.326280299716093e-07, | |
| "loss": 1.6706, | |
| "step": 2164 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.310264543480002e-07, | |
| "loss": 1.6878, | |
| "step": 2166 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.294250564706847e-07, | |
| "loss": 1.692, | |
| "step": 2168 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.278238404664989e-07, | |
| "loss": 1.7141, | |
| "step": 2170 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.262228104618105e-07, | |
| "loss": 1.7044, | |
| "step": 2172 | |
| }, | |
| { | |
| "epoch": 2.15, | |
| "learning_rate": 9.246219705825076e-07, | |
| "loss": 1.6214, | |
| "step": 2174 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.230213249539886e-07, | |
| "loss": 1.7014, | |
| "step": 2176 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.214208777011516e-07, | |
| "loss": 1.7389, | |
| "step": 2178 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.19820632948383e-07, | |
| "loss": 1.731, | |
| "step": 2180 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.18220594819548e-07, | |
| "loss": 1.6979, | |
| "step": 2182 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 9.166207674379783e-07, | |
| "loss": 1.6608, | |
| "step": 2184 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.150211549264631e-07, | |
| "loss": 1.6818, | |
| "step": 2186 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.134217614072378e-07, | |
| "loss": 1.7185, | |
| "step": 2188 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.11822591001974e-07, | |
| "loss": 1.734, | |
| "step": 2190 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.102236478317676e-07, | |
| "loss": 1.6561, | |
| "step": 2192 | |
| }, | |
| { | |
| "epoch": 2.17, | |
| "learning_rate": 9.08624936017129e-07, | |
| "loss": 1.7117, | |
| "step": 2194 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.070264596779725e-07, | |
| "loss": 1.7161, | |
| "step": 2196 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.054282229336055e-07, | |
| "loss": 1.6567, | |
| "step": 2198 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.038302299027188e-07, | |
| "loss": 1.6488, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.022324847033733e-07, | |
| "loss": 1.7681, | |
| "step": 2202 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 9.006349914529932e-07, | |
| "loss": 1.7595, | |
| "step": 2204 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.99037754268352e-07, | |
| "loss": 1.7348, | |
| "step": 2206 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.974407772655641e-07, | |
| "loss": 1.6891, | |
| "step": 2208 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.958440645600739e-07, | |
| "loss": 1.6879, | |
| "step": 2210 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.942476202666426e-07, | |
| "loss": 1.6694, | |
| "step": 2212 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 8.926514484993419e-07, | |
| "loss": 1.6864, | |
| "step": 2214 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.910555533715405e-07, | |
| "loss": 1.6839, | |
| "step": 2216 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.894599389958934e-07, | |
| "loss": 1.7449, | |
| "step": 2218 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.878646094843338e-07, | |
| "loss": 1.6596, | |
| "step": 2220 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.862695689480587e-07, | |
| "loss": 1.6671, | |
| "step": 2222 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 8.84674821497522e-07, | |
| "loss": 1.7066, | |
| "step": 2224 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.830803712424217e-07, | |
| "loss": 1.6791, | |
| "step": 2226 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.8148622229169e-07, | |
| "loss": 1.7459, | |
| "step": 2228 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.798923787534828e-07, | |
| "loss": 1.7514, | |
| "step": 2230 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.782988447351684e-07, | |
| "loss": 1.7402, | |
| "step": 2232 | |
| }, | |
| { | |
| "epoch": 2.21, | |
| "learning_rate": 8.76705624343318e-07, | |
| "loss": 1.7044, | |
| "step": 2234 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.751127216836945e-07, | |
| "loss": 1.6672, | |
| "step": 2236 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.735201408612418e-07, | |
| "loss": 1.7068, | |
| "step": 2238 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.719278859800749e-07, | |
| "loss": 1.6593, | |
| "step": 2240 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.70335961143468e-07, | |
| "loss": 1.6671, | |
| "step": 2242 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 8.687443704538452e-07, | |
| "loss": 1.7018, | |
| "step": 2244 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.671531180127699e-07, | |
| "loss": 1.7241, | |
| "step": 2246 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.655622079209336e-07, | |
| "loss": 1.6957, | |
| "step": 2248 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.639716442781448e-07, | |
| "loss": 1.6789, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.623814311833202e-07, | |
| "loss": 1.6626, | |
| "step": 2252 | |
| }, | |
| { | |
| "epoch": 2.23, | |
| "learning_rate": 8.607915727344727e-07, | |
| "loss": 1.6976, | |
| "step": 2254 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.592020730287013e-07, | |
| "loss": 1.6522, | |
| "step": 2256 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.576129361621808e-07, | |
| "loss": 1.7328, | |
| "step": 2258 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.560241662301504e-07, | |
| "loss": 1.6631, | |
| "step": 2260 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.54435767326904e-07, | |
| "loss": 1.6692, | |
| "step": 2262 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 8.528477435457789e-07, | |
| "loss": 1.6941, | |
| "step": 2264 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.51260098979147e-07, | |
| "loss": 1.6753, | |
| "step": 2266 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.496728377184019e-07, | |
| "loss": 1.685, | |
| "step": 2268 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.480859638539492e-07, | |
| "loss": 1.6844, | |
| "step": 2270 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.464994814751967e-07, | |
| "loss": 1.7154, | |
| "step": 2272 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 8.449133946705433e-07, | |
| "loss": 1.6493, | |
| "step": 2274 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.433277075273684e-07, | |
| "loss": 1.6897, | |
| "step": 2276 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.417424241320219e-07, | |
| "loss": 1.7593, | |
| "step": 2278 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.401575485698118e-07, | |
| "loss": 1.7086, | |
| "step": 2280 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.385730849249969e-07, | |
| "loss": 1.7061, | |
| "step": 2282 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 8.369890372807736e-07, | |
| "loss": 1.7041, | |
| "step": 2284 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.354054097192659e-07, | |
| "loss": 1.7031, | |
| "step": 2286 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.338222063215168e-07, | |
| "loss": 1.6871, | |
| "step": 2288 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.322394311674735e-07, | |
| "loss": 1.7028, | |
| "step": 2290 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.306570883359824e-07, | |
| "loss": 1.6795, | |
| "step": 2292 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.290751819047743e-07, | |
| "loss": 1.6632, | |
| "step": 2294 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 8.274937159504558e-07, | |
| "loss": 1.7115, | |
| "step": 2296 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.25912694548498e-07, | |
| "loss": 1.6845, | |
| "step": 2298 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.243321217732267e-07, | |
| "loss": 1.6457, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.227520016978117e-07, | |
| "loss": 1.6617, | |
| "step": 2302 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.211723383942559e-07, | |
| "loss": 1.6927, | |
| "step": 2304 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 8.195931359333857e-07, | |
| "loss": 1.7118, | |
| "step": 2306 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.180143983848387e-07, | |
| "loss": 1.6656, | |
| "step": 2308 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.164361298170553e-07, | |
| "loss": 1.6978, | |
| "step": 2310 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.148583342972671e-07, | |
| "loss": 1.6868, | |
| "step": 2312 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.132810158914871e-07, | |
| "loss": 1.6648, | |
| "step": 2314 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 8.117041786644983e-07, | |
| "loss": 1.682, | |
| "step": 2316 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.109159417629565e-07, | |
| "loss": 1.7076, | |
| "step": 2318 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.101278266798435e-07, | |
| "loss": 1.6716, | |
| "step": 2320 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.08551963999815e-07, | |
| "loss": 1.7117, | |
| "step": 2322 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.069765946854449e-07, | |
| "loss": 1.6857, | |
| "step": 2324 | |
| }, | |
| { | |
| "epoch": 2.3, | |
| "learning_rate": 8.054017227964933e-07, | |
| "loss": 1.7236, | |
| "step": 2326 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 8.038273523914387e-07, | |
| "loss": 1.6899, | |
| "step": 2328 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 8.022534875274667e-07, | |
| "loss": 1.7217, | |
| "step": 2330 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 8.006801322604609e-07, | |
| "loss": 1.6659, | |
| "step": 2332 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 7.991072906449905e-07, | |
| "loss": 1.6978, | |
| "step": 2334 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 7.975349667343028e-07, | |
| "loss": 1.6926, | |
| "step": 2336 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.959631645803097e-07, | |
| "loss": 1.7067, | |
| "step": 2338 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.943918882335783e-07, | |
| "loss": 1.6996, | |
| "step": 2340 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.928211417433214e-07, | |
| "loss": 1.688, | |
| "step": 2342 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.912509291573858e-07, | |
| "loss": 1.7048, | |
| "step": 2344 | |
| }, | |
| { | |
| "epoch": 2.32, | |
| "learning_rate": 7.896812545222438e-07, | |
| "loss": 1.7409, | |
| "step": 2346 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.881121218829787e-07, | |
| "loss": 1.675, | |
| "step": 2348 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.865435352832799e-07, | |
| "loss": 1.7349, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.849754987654278e-07, | |
| "loss": 1.7014, | |
| "step": 2352 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.834080163702861e-07, | |
| "loss": 1.6584, | |
| "step": 2354 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 7.818410921372903e-07, | |
| "loss": 1.7162, | |
| "step": 2356 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.802747301044371e-07, | |
| "loss": 1.6483, | |
| "step": 2358 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.787089343082748e-07, | |
| "loss": 1.7095, | |
| "step": 2360 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.771437087838929e-07, | |
| "loss": 1.6742, | |
| "step": 2362 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.755790575649105e-07, | |
| "loss": 1.6931, | |
| "step": 2364 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 7.740149846834673e-07, | |
| "loss": 1.6985, | |
| "step": 2366 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.724514941702121e-07, | |
| "loss": 1.7498, | |
| "step": 2368 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.708885900542929e-07, | |
| "loss": 1.7108, | |
| "step": 2370 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.693262763633477e-07, | |
| "loss": 1.7009, | |
| "step": 2372 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.677645571234913e-07, | |
| "loss": 1.6861, | |
| "step": 2374 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 7.662034363593079e-07, | |
| "loss": 1.6602, | |
| "step": 2376 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.646429180938386e-07, | |
| "loss": 1.6654, | |
| "step": 2378 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.630830063485719e-07, | |
| "loss": 1.7435, | |
| "step": 2380 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.615237051434341e-07, | |
| "loss": 1.7151, | |
| "step": 2382 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.599650184967772e-07, | |
| "loss": 1.6959, | |
| "step": 2384 | |
| }, | |
| { | |
| "epoch": 2.36, | |
| "learning_rate": 7.584069504253701e-07, | |
| "loss": 1.7323, | |
| "step": 2386 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.56849504944387e-07, | |
| "loss": 1.7185, | |
| "step": 2388 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.552926860673979e-07, | |
| "loss": 1.7051, | |
| "step": 2390 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.537364978063587e-07, | |
| "loss": 1.6372, | |
| "step": 2392 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.521809441715994e-07, | |
| "loss": 1.6673, | |
| "step": 2394 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 7.506260291718145e-07, | |
| "loss": 1.6899, | |
| "step": 2396 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.490717568140532e-07, | |
| "loss": 1.7418, | |
| "step": 2398 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.475181311037084e-07, | |
| "loss": 1.6876, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.459651560445063e-07, | |
| "loss": 1.6791, | |
| "step": 2402 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.444128356384978e-07, | |
| "loss": 1.7036, | |
| "step": 2404 | |
| }, | |
| { | |
| "epoch": 2.38, | |
| "learning_rate": 7.42861173886044e-07, | |
| "loss": 1.7317, | |
| "step": 2406 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.413101747858111e-07, | |
| "loss": 1.7448, | |
| "step": 2408 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.397598423347564e-07, | |
| "loss": 1.6793, | |
| "step": 2410 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.382101805281198e-07, | |
| "loss": 1.649, | |
| "step": 2412 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.366611933594131e-07, | |
| "loss": 1.7431, | |
| "step": 2414 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 7.351128848204083e-07, | |
| "loss": 1.692, | |
| "step": 2416 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.335652589011299e-07, | |
| "loss": 1.7502, | |
| "step": 2418 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.320183195898428e-07, | |
| "loss": 1.6684, | |
| "step": 2420 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.304720708730422e-07, | |
| "loss": 1.6798, | |
| "step": 2422 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.289265167354448e-07, | |
| "loss": 1.6456, | |
| "step": 2424 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 7.273816611599751e-07, | |
| "loss": 1.6918, | |
| "step": 2426 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.258375081277597e-07, | |
| "loss": 1.7056, | |
| "step": 2428 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.242940616181136e-07, | |
| "loss": 1.7111, | |
| "step": 2430 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.227513256085312e-07, | |
| "loss": 1.6859, | |
| "step": 2432 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.212093040746762e-07, | |
| "loss": 1.7127, | |
| "step": 2434 | |
| }, | |
| { | |
| "epoch": 2.41, | |
| "learning_rate": 7.196680009903701e-07, | |
| "loss": 1.7154, | |
| "step": 2436 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.181274203275847e-07, | |
| "loss": 1.6469, | |
| "step": 2438 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.165875660564286e-07, | |
| "loss": 1.7553, | |
| "step": 2440 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.150484421451393e-07, | |
| "loss": 1.6825, | |
| "step": 2442 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.135100525600715e-07, | |
| "loss": 1.6694, | |
| "step": 2444 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 7.119724012656878e-07, | |
| "loss": 1.6641, | |
| "step": 2446 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.104354922245483e-07, | |
| "loss": 1.6566, | |
| "step": 2448 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.088993293973008e-07, | |
| "loss": 1.7227, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.073639167426692e-07, | |
| "loss": 1.7445, | |
| "step": 2452 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.058292582174444e-07, | |
| "loss": 1.7304, | |
| "step": 2454 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 7.042953577764741e-07, | |
| "loss": 1.7014, | |
| "step": 2456 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 7.027622193726524e-07, | |
| "loss": 1.6649, | |
| "step": 2458 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 7.012298469569095e-07, | |
| "loss": 1.6338, | |
| "step": 2460 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 6.99698244478202e-07, | |
| "loss": 1.6842, | |
| "step": 2462 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 6.981674158835018e-07, | |
| "loss": 1.6714, | |
| "step": 2464 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 6.966373651177865e-07, | |
| "loss": 1.7427, | |
| "step": 2466 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.951080961240296e-07, | |
| "loss": 1.6865, | |
| "step": 2468 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.935796128431901e-07, | |
| "loss": 1.7064, | |
| "step": 2470 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.920519192142018e-07, | |
| "loss": 1.6265, | |
| "step": 2472 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.905250191739633e-07, | |
| "loss": 1.6545, | |
| "step": 2474 | |
| }, | |
| { | |
| "epoch": 2.45, | |
| "learning_rate": 6.889989166573287e-07, | |
| "loss": 1.7049, | |
| "step": 2476 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.874736155970967e-07, | |
| "loss": 1.7092, | |
| "step": 2478 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.859491199240001e-07, | |
| "loss": 1.7326, | |
| "step": 2480 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.844254335666979e-07, | |
| "loss": 1.6849, | |
| "step": 2482 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.829025604517606e-07, | |
| "loss": 1.696, | |
| "step": 2484 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 6.813805045036655e-07, | |
| "loss": 1.7074, | |
| "step": 2486 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.798592696447832e-07, | |
| "loss": 1.7195, | |
| "step": 2488 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.783388597953681e-07, | |
| "loss": 1.642, | |
| "step": 2490 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.768192788735484e-07, | |
| "loss": 1.6769, | |
| "step": 2492 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.753005307953165e-07, | |
| "loss": 1.7308, | |
| "step": 2494 | |
| }, | |
| { | |
| "epoch": 2.47, | |
| "learning_rate": 6.737826194745188e-07, | |
| "loss": 1.6753, | |
| "step": 2496 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.72265548822845e-07, | |
| "loss": 1.7083, | |
| "step": 2498 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.707493227498186e-07, | |
| "loss": 1.6606, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.692339451627858e-07, | |
| "loss": 1.6709, | |
| "step": 2502 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.677194199669071e-07, | |
| "loss": 1.6954, | |
| "step": 2504 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 6.662057510651463e-07, | |
| "loss": 1.6816, | |
| "step": 2506 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.646929423582603e-07, | |
| "loss": 1.7113, | |
| "step": 2508 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.631809977447896e-07, | |
| "loss": 1.6677, | |
| "step": 2510 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.616699211210468e-07, | |
| "loss": 1.7202, | |
| "step": 2512 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.601597163811087e-07, | |
| "loss": 1.6861, | |
| "step": 2514 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.586503874168057e-07, | |
| "loss": 1.7311, | |
| "step": 2516 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 6.5714193811771e-07, | |
| "loss": 1.6645, | |
| "step": 2518 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.55634372371128e-07, | |
| "loss": 1.6595, | |
| "step": 2520 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.541276940620879e-07, | |
| "loss": 1.6938, | |
| "step": 2522 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.526219070733325e-07, | |
| "loss": 1.7347, | |
| "step": 2524 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.511170152853062e-07, | |
| "loss": 1.7152, | |
| "step": 2526 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 6.496130225761476e-07, | |
| "loss": 1.6773, | |
| "step": 2528 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.481099328216781e-07, | |
| "loss": 1.6889, | |
| "step": 2530 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.466077498953915e-07, | |
| "loss": 1.7002, | |
| "step": 2532 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.451064776684449e-07, | |
| "loss": 1.697, | |
| "step": 2534 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.436061200096492e-07, | |
| "loss": 1.7142, | |
| "step": 2536 | |
| }, | |
| { | |
| "epoch": 2.51, | |
| "learning_rate": 6.421066807854584e-07, | |
| "loss": 1.687, | |
| "step": 2538 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.406081638599581e-07, | |
| "loss": 1.6856, | |
| "step": 2540 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.391105730948591e-07, | |
| "loss": 1.7328, | |
| "step": 2542 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.376139123494843e-07, | |
| "loss": 1.7172, | |
| "step": 2544 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.361181854807604e-07, | |
| "loss": 1.7127, | |
| "step": 2546 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 6.346233963432077e-07, | |
| "loss": 1.6909, | |
| "step": 2548 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.331295487889288e-07, | |
| "loss": 1.6796, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.316366466676011e-07, | |
| "loss": 1.6829, | |
| "step": 2552 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.301446938264648e-07, | |
| "loss": 1.6658, | |
| "step": 2554 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.286536941103146e-07, | |
| "loss": 1.7522, | |
| "step": 2556 | |
| }, | |
| { | |
| "epoch": 2.53, | |
| "learning_rate": 6.271636513614882e-07, | |
| "loss": 1.691, | |
| "step": 2558 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.256745694198568e-07, | |
| "loss": 1.704, | |
| "step": 2560 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.241864521228173e-07, | |
| "loss": 1.7034, | |
| "step": 2562 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.226993033052789e-07, | |
| "loss": 1.7227, | |
| "step": 2564 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.21213126799656e-07, | |
| "loss": 1.7184, | |
| "step": 2566 | |
| }, | |
| { | |
| "epoch": 2.54, | |
| "learning_rate": 6.197279264358572e-07, | |
| "loss": 1.737, | |
| "step": 2568 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.182437060412749e-07, | |
| "loss": 1.6819, | |
| "step": 2570 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.167604694407766e-07, | |
| "loss": 1.7859, | |
| "step": 2572 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.152782204566951e-07, | |
| "loss": 1.679, | |
| "step": 2574 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.137969629088174e-07, | |
| "loss": 1.6912, | |
| "step": 2576 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 6.123167006143755e-07, | |
| "loss": 1.688, | |
| "step": 2578 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.108374373880367e-07, | |
| "loss": 1.6896, | |
| "step": 2580 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.093591770418938e-07, | |
| "loss": 1.7549, | |
| "step": 2582 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.078819233854555e-07, | |
| "loss": 1.7085, | |
| "step": 2584 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.064056802256358e-07, | |
| "loss": 1.6606, | |
| "step": 2586 | |
| }, | |
| { | |
| "epoch": 2.56, | |
| "learning_rate": 6.049304513667442e-07, | |
| "loss": 1.6915, | |
| "step": 2588 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 6.034562406104773e-07, | |
| "loss": 1.6713, | |
| "step": 2590 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 6.019830517559073e-07, | |
| "loss": 1.6162, | |
| "step": 2592 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 6.005108885994732e-07, | |
| "loss": 1.6848, | |
| "step": 2594 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 5.990397549349713e-07, | |
| "loss": 1.7615, | |
| "step": 2596 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 5.975696545535436e-07, | |
| "loss": 1.7274, | |
| "step": 2598 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.961005912436703e-07, | |
| "loss": 1.6831, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.946325687911588e-07, | |
| "loss": 1.7243, | |
| "step": 2602 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.93165590979134e-07, | |
| "loss": 1.6636, | |
| "step": 2604 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.916996615880295e-07, | |
| "loss": 1.7544, | |
| "step": 2606 | |
| }, | |
| { | |
| "epoch": 2.58, | |
| "learning_rate": 5.902347843955758e-07, | |
| "loss": 1.7464, | |
| "step": 2608 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.887709631767927e-07, | |
| "loss": 1.709, | |
| "step": 2610 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.87308201703979e-07, | |
| "loss": 1.6589, | |
| "step": 2612 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.858465037467014e-07, | |
| "loss": 1.7203, | |
| "step": 2614 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.843858730717877e-07, | |
| "loss": 1.6294, | |
| "step": 2616 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 5.829263134433129e-07, | |
| "loss": 1.7211, | |
| "step": 2618 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.814678286225938e-07, | |
| "loss": 1.6945, | |
| "step": 2620 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.800104223681772e-07, | |
| "loss": 1.6875, | |
| "step": 2622 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.785540984358293e-07, | |
| "loss": 1.7123, | |
| "step": 2624 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.770988605785285e-07, | |
| "loss": 1.6918, | |
| "step": 2626 | |
| }, | |
| { | |
| "epoch": 2.6, | |
| "learning_rate": 5.756447125464532e-07, | |
| "loss": 1.6703, | |
| "step": 2628 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.741916580869736e-07, | |
| "loss": 1.6663, | |
| "step": 2630 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.727397009446422e-07, | |
| "loss": 1.6892, | |
| "step": 2632 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.712888448611843e-07, | |
| "loss": 1.7073, | |
| "step": 2634 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.698390935754853e-07, | |
| "loss": 1.7169, | |
| "step": 2636 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 5.683904508235857e-07, | |
| "loss": 1.71, | |
| "step": 2638 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.66942920338669e-07, | |
| "loss": 1.7353, | |
| "step": 2640 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.654965058510516e-07, | |
| "loss": 1.6847, | |
| "step": 2642 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.64051211088175e-07, | |
| "loss": 1.7284, | |
| "step": 2644 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.626070397745937e-07, | |
| "loss": 1.6899, | |
| "step": 2646 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 5.611639956319687e-07, | |
| "loss": 1.7052, | |
| "step": 2648 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.597220823790545e-07, | |
| "loss": 1.6736, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.582813037316926e-07, | |
| "loss": 1.6843, | |
| "step": 2652 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.568416634028013e-07, | |
| "loss": 1.6836, | |
| "step": 2654 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.554031651023625e-07, | |
| "loss": 1.643, | |
| "step": 2656 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 5.539658125374177e-07, | |
| "loss": 1.7168, | |
| "step": 2658 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.525296094120555e-07, | |
| "loss": 1.6669, | |
| "step": 2660 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.51094559427401e-07, | |
| "loss": 1.7637, | |
| "step": 2662 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.496606662816093e-07, | |
| "loss": 1.6912, | |
| "step": 2664 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.482279336698527e-07, | |
| "loss": 1.7166, | |
| "step": 2666 | |
| }, | |
| { | |
| "epoch": 2.64, | |
| "learning_rate": 5.467963652843146e-07, | |
| "loss": 1.7496, | |
| "step": 2668 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.453659648141762e-07, | |
| "loss": 1.6845, | |
| "step": 2670 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.439367359456106e-07, | |
| "loss": 1.6562, | |
| "step": 2672 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.425086823617708e-07, | |
| "loss": 1.6745, | |
| "step": 2674 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.410818077427807e-07, | |
| "loss": 1.7184, | |
| "step": 2676 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 5.396561157657271e-07, | |
| "loss": 1.6447, | |
| "step": 2678 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.382316101046487e-07, | |
| "loss": 1.731, | |
| "step": 2680 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.368082944305265e-07, | |
| "loss": 1.6807, | |
| "step": 2682 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.35386172411275e-07, | |
| "loss": 1.7142, | |
| "step": 2684 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.339652477117335e-07, | |
| "loss": 1.6614, | |
| "step": 2686 | |
| }, | |
| { | |
| "epoch": 2.66, | |
| "learning_rate": 5.325455239936545e-07, | |
| "loss": 1.6886, | |
| "step": 2688 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.311270049156966e-07, | |
| "loss": 1.6559, | |
| "step": 2690 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.297096941334138e-07, | |
| "loss": 1.7406, | |
| "step": 2692 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.28293595299246e-07, | |
| "loss": 1.6698, | |
| "step": 2694 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.268787120625096e-07, | |
| "loss": 1.6907, | |
| "step": 2696 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 5.254650480693888e-07, | |
| "loss": 1.6777, | |
| "step": 2698 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.240526069629264e-07, | |
| "loss": 1.6978, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.226413923830128e-07, | |
| "loss": 1.7113, | |
| "step": 2702 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.212314079663777e-07, | |
| "loss": 1.64, | |
| "step": 2704 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.198226573465813e-07, | |
| "loss": 1.6904, | |
| "step": 2706 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 5.184151441540035e-07, | |
| "loss": 1.6339, | |
| "step": 2708 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.170088720158359e-07, | |
| "loss": 1.6761, | |
| "step": 2710 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.156038445560724e-07, | |
| "loss": 1.7021, | |
| "step": 2712 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.142000653954973e-07, | |
| "loss": 1.7174, | |
| "step": 2714 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.1279753815168e-07, | |
| "loss": 1.6789, | |
| "step": 2716 | |
| }, | |
| { | |
| "epoch": 2.69, | |
| "learning_rate": 5.113962664389635e-07, | |
| "loss": 1.6851, | |
| "step": 2718 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.099962538684541e-07, | |
| "loss": 1.7208, | |
| "step": 2720 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.085975040480139e-07, | |
| "loss": 1.6654, | |
| "step": 2722 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.072000205822514e-07, | |
| "loss": 1.6596, | |
| "step": 2724 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.05803807072511e-07, | |
| "loss": 1.6819, | |
| "step": 2726 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.044088671168644e-07, | |
| "loss": 1.7198, | |
| "step": 2728 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 5.030152043101019e-07, | |
| "loss": 1.6521, | |
| "step": 2730 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 5.016228222437222e-07, | |
| "loss": 1.7227, | |
| "step": 2732 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 5.002317245059227e-07, | |
| "loss": 1.7101, | |
| "step": 2734 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.988419146815926e-07, | |
| "loss": 1.6724, | |
| "step": 2736 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.974533963523014e-07, | |
| "loss": 1.7026, | |
| "step": 2738 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 4.9606617309629e-07, | |
| "loss": 1.6954, | |
| "step": 2740 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.94680248488462e-07, | |
| "loss": 1.6506, | |
| "step": 2742 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.932956261003749e-07, | |
| "loss": 1.645, | |
| "step": 2744 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.919123095002304e-07, | |
| "loss": 1.6815, | |
| "step": 2746 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.905303022528641e-07, | |
| "loss": 1.6839, | |
| "step": 2748 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 4.891496079197387e-07, | |
| "loss": 1.6525, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.877702300589324e-07, | |
| "loss": 1.7223, | |
| "step": 2752 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.863921722251312e-07, | |
| "loss": 1.7158, | |
| "step": 2754 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.850154379696194e-07, | |
| "loss": 1.7095, | |
| "step": 2756 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.836400308402711e-07, | |
| "loss": 1.6946, | |
| "step": 2758 | |
| }, | |
| { | |
| "epoch": 2.73, | |
| "learning_rate": 4.822659543815393e-07, | |
| "loss": 1.6986, | |
| "step": 2760 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.808932121344478e-07, | |
| "loss": 1.6826, | |
| "step": 2762 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.795218076365832e-07, | |
| "loss": 1.6625, | |
| "step": 2764 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.781517444220835e-07, | |
| "loss": 1.6878, | |
| "step": 2766 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.7678302602163135e-07, | |
| "loss": 1.6795, | |
| "step": 2768 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.754156559624425e-07, | |
| "loss": 1.7281, | |
| "step": 2770 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.7404963776825925e-07, | |
| "loss": 1.7138, | |
| "step": 2772 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.72684974959339e-07, | |
| "loss": 1.6189, | |
| "step": 2774 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.713216710524469e-07, | |
| "loss": 1.7349, | |
| "step": 2776 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.6995972956084717e-07, | |
| "loss": 1.7164, | |
| "step": 2778 | |
| }, | |
| { | |
| "epoch": 2.75, | |
| "learning_rate": 4.685991539942903e-07, | |
| "loss": 1.6753, | |
| "step": 2780 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.672399478590092e-07, | |
| "loss": 1.6889, | |
| "step": 2782 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.6588211465770735e-07, | |
| "loss": 1.7264, | |
| "step": 2784 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.645256578895491e-07, | |
| "loss": 1.6632, | |
| "step": 2786 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.6317058105015306e-07, | |
| "loss": 1.6736, | |
| "step": 2788 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 4.6181688763158024e-07, | |
| "loss": 1.721, | |
| "step": 2790 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.604645811223282e-07, | |
| "loss": 1.6435, | |
| "step": 2792 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.5911366500731873e-07, | |
| "loss": 1.7203, | |
| "step": 2794 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.577641427678923e-07, | |
| "loss": 1.6521, | |
| "step": 2796 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.564160178817958e-07, | |
| "loss": 1.6348, | |
| "step": 2798 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 4.550692938231755e-07, | |
| "loss": 1.643, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.5372397406256824e-07, | |
| "loss": 1.6576, | |
| "step": 2802 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.5238006206689204e-07, | |
| "loss": 1.7095, | |
| "step": 2804 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.5103756129943593e-07, | |
| "loss": 1.6866, | |
| "step": 2806 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.4969647521985363e-07, | |
| "loss": 1.6594, | |
| "step": 2808 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 4.4835680728415217e-07, | |
| "loss": 1.7039, | |
| "step": 2810 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.4701856094468383e-07, | |
| "loss": 1.6914, | |
| "step": 2812 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.4568173965013835e-07, | |
| "loss": 1.6774, | |
| "step": 2814 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.4434634684553286e-07, | |
| "loss": 1.6971, | |
| "step": 2816 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.430123859722027e-07, | |
| "loss": 1.724, | |
| "step": 2818 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 4.4167986046779315e-07, | |
| "loss": 1.6961, | |
| "step": 2820 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.403487737662509e-07, | |
| "loss": 1.7321, | |
| "step": 2822 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.3901912929781524e-07, | |
| "loss": 1.6939, | |
| "step": 2824 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.376909304890077e-07, | |
| "loss": 1.6564, | |
| "step": 2826 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.3636418076262473e-07, | |
| "loss": 1.628, | |
| "step": 2828 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 4.35038883537729e-07, | |
| "loss": 1.6867, | |
| "step": 2830 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.3371504222963906e-07, | |
| "loss": 1.7011, | |
| "step": 2832 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.323926602499224e-07, | |
| "loss": 1.6426, | |
| "step": 2834 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.31071741006386e-07, | |
| "loss": 1.6692, | |
| "step": 2836 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.2975228790306627e-07, | |
| "loss": 1.6933, | |
| "step": 2838 | |
| }, | |
| { | |
| "epoch": 2.81, | |
| "learning_rate": 4.2843430434022166e-07, | |
| "loss": 1.7154, | |
| "step": 2840 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.2711779371432445e-07, | |
| "loss": 1.739, | |
| "step": 2842 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.258027594180499e-07, | |
| "loss": 1.6868, | |
| "step": 2844 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.2448920484027006e-07, | |
| "loss": 1.7613, | |
| "step": 2846 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.231771333660422e-07, | |
| "loss": 1.6815, | |
| "step": 2848 | |
| }, | |
| { | |
| "epoch": 2.82, | |
| "learning_rate": 4.2186654837660316e-07, | |
| "loss": 1.6944, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.2055745324935763e-07, | |
| "loss": 1.6703, | |
| "step": 2852 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.1924985135787194e-07, | |
| "loss": 1.6135, | |
| "step": 2854 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.1794374607186453e-07, | |
| "loss": 1.6998, | |
| "step": 2856 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.166391407571952e-07, | |
| "loss": 1.7389, | |
| "step": 2858 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 4.153360387758603e-07, | |
| "loss": 1.7111, | |
| "step": 2860 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.140344434859816e-07, | |
| "loss": 1.6901, | |
| "step": 2862 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.1273435824179757e-07, | |
| "loss": 1.7234, | |
| "step": 2864 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.1143578639365507e-07, | |
| "loss": 1.6779, | |
| "step": 2866 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.101387312880018e-07, | |
| "loss": 1.7124, | |
| "step": 2868 | |
| }, | |
| { | |
| "epoch": 2.84, | |
| "learning_rate": 4.0884319626737664e-07, | |
| "loss": 1.6751, | |
| "step": 2870 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.0754918467040013e-07, | |
| "loss": 1.6454, | |
| "step": 2872 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.062566998317685e-07, | |
| "loss": 1.7012, | |
| "step": 2874 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.049657450822421e-07, | |
| "loss": 1.684, | |
| "step": 2876 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.0367632374863856e-07, | |
| "loss": 1.7018, | |
| "step": 2878 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 4.0238843915382435e-07, | |
| "loss": 1.682, | |
| "step": 2880 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 4.01102094616706e-07, | |
| "loss": 1.7358, | |
| "step": 2882 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.9981729345222005e-07, | |
| "loss": 1.6558, | |
| "step": 2884 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.985340389713262e-07, | |
| "loss": 1.7431, | |
| "step": 2886 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.972523344809993e-07, | |
| "loss": 1.6865, | |
| "step": 2888 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 3.959721832842182e-07, | |
| "loss": 1.7101, | |
| "step": 2890 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.946935886799603e-07, | |
| "loss": 1.6728, | |
| "step": 2892 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.934165539631913e-07, | |
| "loss": 1.6375, | |
| "step": 2894 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.9214108242485645e-07, | |
| "loss": 1.7287, | |
| "step": 2896 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.908671773518727e-07, | |
| "loss": 1.6645, | |
| "step": 2898 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 3.895948420271209e-07, | |
| "loss": 1.7205, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.883240797294365e-07, | |
| "loss": 1.6477, | |
| "step": 2902 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.8705489373360055e-07, | |
| "loss": 1.6981, | |
| "step": 2904 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.8578728731033214e-07, | |
| "loss": 1.7069, | |
| "step": 2906 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.845212637262806e-07, | |
| "loss": 1.7126, | |
| "step": 2908 | |
| }, | |
| { | |
| "epoch": 2.88, | |
| "learning_rate": 3.8325682624401487e-07, | |
| "loss": 1.661, | |
| "step": 2910 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.8199397812201803e-07, | |
| "loss": 1.6845, | |
| "step": 2912 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.807327226146757e-07, | |
| "loss": 1.6835, | |
| "step": 2914 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.79473062972271e-07, | |
| "loss": 1.7238, | |
| "step": 2916 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.782150024409727e-07, | |
| "loss": 1.6807, | |
| "step": 2918 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 3.769585442628304e-07, | |
| "loss": 1.7436, | |
| "step": 2920 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.757036916757631e-07, | |
| "loss": 1.6375, | |
| "step": 2922 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.744504479135522e-07, | |
| "loss": 1.7159, | |
| "step": 2924 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.7319881620583396e-07, | |
| "loss": 1.6949, | |
| "step": 2926 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.719487997780902e-07, | |
| "loss": 1.6679, | |
| "step": 2928 | |
| }, | |
| { | |
| "epoch": 2.9, | |
| "learning_rate": 3.707004018516392e-07, | |
| "loss": 1.6609, | |
| "step": 2930 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.6945362564362946e-07, | |
| "loss": 1.7207, | |
| "step": 2932 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.682084743670296e-07, | |
| "loss": 1.7001, | |
| "step": 2934 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.6696495123062056e-07, | |
| "loss": 1.69, | |
| "step": 2936 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.6572305943898817e-07, | |
| "loss": 1.6587, | |
| "step": 2938 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 3.644828021925144e-07, | |
| "loss": 1.7012, | |
| "step": 2940 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.6324418268736813e-07, | |
| "loss": 1.6881, | |
| "step": 2942 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.620072041154978e-07, | |
| "loss": 1.6681, | |
| "step": 2944 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.607718696646238e-07, | |
| "loss": 1.6934, | |
| "step": 2946 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.5953818251822976e-07, | |
| "loss": 1.6946, | |
| "step": 2948 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.5830614585555296e-07, | |
| "loss": 1.6701, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 3.570757628515786e-07, | |
| "loss": 1.625, | |
| "step": 2952 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.558470366770295e-07, | |
| "loss": 1.6427, | |
| "step": 2954 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.546199704983591e-07, | |
| "loss": 1.7193, | |
| "step": 2956 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.53394567477743e-07, | |
| "loss": 1.6401, | |
| "step": 2958 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.521708307730713e-07, | |
| "loss": 1.7526, | |
| "step": 2960 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 3.509487635379392e-07, | |
| "loss": 1.6339, | |
| "step": 2962 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.497283689216397e-07, | |
| "loss": 1.6991, | |
| "step": 2964 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.485096500691561e-07, | |
| "loss": 1.6556, | |
| "step": 2966 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.472926101211522e-07, | |
| "loss": 1.7234, | |
| "step": 2968 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.4607725221396653e-07, | |
| "loss": 1.6865, | |
| "step": 2970 | |
| }, | |
| { | |
| "epoch": 2.94, | |
| "learning_rate": 3.4486357947960164e-07, | |
| "loss": 1.7662, | |
| "step": 2972 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.436515950457186e-07, | |
| "loss": 1.7085, | |
| "step": 2974 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.4244130203562626e-07, | |
| "loss": 1.6606, | |
| "step": 2976 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.4123270356827594e-07, | |
| "loss": 1.6801, | |
| "step": 2978 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.400258027582522e-07, | |
| "loss": 1.6536, | |
| "step": 2980 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 3.388206027157626e-07, | |
| "loss": 1.6931, | |
| "step": 2982 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.376171065466341e-07, | |
| "loss": 1.6881, | |
| "step": 2984 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.3641531735230233e-07, | |
| "loss": 1.7132, | |
| "step": 2986 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.352152382298027e-07, | |
| "loss": 1.6624, | |
| "step": 2988 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.3401687227176534e-07, | |
| "loss": 1.6712, | |
| "step": 2990 | |
| }, | |
| { | |
| "epoch": 2.96, | |
| "learning_rate": 3.328202225664042e-07, | |
| "loss": 1.6864, | |
| "step": 2992 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.3162529219751155e-07, | |
| "loss": 1.6854, | |
| "step": 2994 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.304320842444479e-07, | |
| "loss": 1.7078, | |
| "step": 2996 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.29240601782136e-07, | |
| "loss": 1.7342, | |
| "step": 2998 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.280508478810512e-07, | |
| "loss": 1.7156, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 2.97, | |
| "learning_rate": 3.2686282560721424e-07, | |
| "loss": 1.6919, | |
| "step": 3002 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.256765380221841e-07, | |
| "loss": 1.6578, | |
| "step": 3004 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.2449198818304956e-07, | |
| "loss": 1.7127, | |
| "step": 3006 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.2330917914242016e-07, | |
| "loss": 1.7205, | |
| "step": 3008 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.221281139484199e-07, | |
| "loss": 1.7033, | |
| "step": 3010 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 3.2094879564467926e-07, | |
| "loss": 1.7026, | |
| "step": 3012 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.1977122727032613e-07, | |
| "loss": 1.6754, | |
| "step": 3014 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.185954118599795e-07, | |
| "loss": 1.6521, | |
| "step": 3016 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.17421352443741e-07, | |
| "loss": 1.6882, | |
| "step": 3018 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.1624905204718645e-07, | |
| "loss": 1.7017, | |
| "step": 3020 | |
| }, | |
| { | |
| "epoch": 2.99, | |
| "learning_rate": 3.150785136913584e-07, | |
| "loss": 1.7032, | |
| "step": 3022 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.1390974039275943e-07, | |
| "loss": 1.6883, | |
| "step": 3024 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.127427351633437e-07, | |
| "loss": 1.7084, | |
| "step": 3026 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.1157750101050795e-07, | |
| "loss": 1.6742, | |
| "step": 3028 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.104140409370851e-07, | |
| "loss": 1.623, | |
| "step": 3030 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 3.0925235794133717e-07, | |
| "loss": 1.6781, | |
| "step": 3032 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.080924550169453e-07, | |
| "loss": 1.6811, | |
| "step": 3034 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.069343351530043e-07, | |
| "loss": 1.6752, | |
| "step": 3036 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.0577800133401387e-07, | |
| "loss": 1.7036, | |
| "step": 3038 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.0462345653987066e-07, | |
| "loss": 1.6701, | |
| "step": 3040 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 3.034707037458608e-07, | |
| "loss": 1.692, | |
| "step": 3042 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 3.023197459226534e-07, | |
| "loss": 1.6856, | |
| "step": 3044 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 3.0117058603629054e-07, | |
| "loss": 1.6747, | |
| "step": 3046 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 3.0002322704818237e-07, | |
| "loss": 1.7099, | |
| "step": 3048 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 2.988776719150967e-07, | |
| "loss": 1.6694, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 2.977339235891541e-07, | |
| "loss": 1.6709, | |
| "step": 3052 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.965919850178177e-07, | |
| "loss": 1.6794, | |
| "step": 3054 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.9545185914388817e-07, | |
| "loss": 1.6777, | |
| "step": 3056 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.943135489054936e-07, | |
| "loss": 1.641, | |
| "step": 3058 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.9317705723608365e-07, | |
| "loss": 1.6374, | |
| "step": 3060 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 2.920423870644216e-07, | |
| "loss": 1.6835, | |
| "step": 3062 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.90909541314577e-07, | |
| "loss": 1.6734, | |
| "step": 3064 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.8977852290591707e-07, | |
| "loss": 1.6777, | |
| "step": 3066 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.8864933475310006e-07, | |
| "loss": 1.7276, | |
| "step": 3068 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.875219797660681e-07, | |
| "loss": 1.7273, | |
| "step": 3070 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 2.863964608500392e-07, | |
| "loss": 1.6844, | |
| "step": 3072 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.85272780905499e-07, | |
| "loss": 1.7278, | |
| "step": 3074 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.841509428281953e-07, | |
| "loss": 1.6933, | |
| "step": 3076 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.830309495091282e-07, | |
| "loss": 1.6473, | |
| "step": 3078 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.8191280383454394e-07, | |
| "loss": 1.7405, | |
| "step": 3080 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 2.807965086859283e-07, | |
| "loss": 1.6804, | |
| "step": 3082 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.796820669399976e-07, | |
| "loss": 1.6326, | |
| "step": 3084 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.7856948146869177e-07, | |
| "loss": 1.754, | |
| "step": 3086 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.774587551391667e-07, | |
| "loss": 1.6694, | |
| "step": 3088 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.763498908137887e-07, | |
| "loss": 1.72, | |
| "step": 3090 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 2.752428913501239e-07, | |
| "loss": 1.6509, | |
| "step": 3092 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.741377596009338e-07, | |
| "loss": 1.6915, | |
| "step": 3094 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.7303449841416657e-07, | |
| "loss": 1.7006, | |
| "step": 3096 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.7193311063294964e-07, | |
| "loss": 1.6722, | |
| "step": 3098 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.7083359909558235e-07, | |
| "loss": 1.6992, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 2.6973596663552957e-07, | |
| "loss": 1.6196, | |
| "step": 3102 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.686402160814141e-07, | |
| "loss": 1.6996, | |
| "step": 3104 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.6754635025700724e-07, | |
| "loss": 1.7231, | |
| "step": 3106 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.66454371981225e-07, | |
| "loss": 1.6904, | |
| "step": 3108 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.6536428406811904e-07, | |
| "loss": 1.6754, | |
| "step": 3110 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 2.642760893268684e-07, | |
| "loss": 1.6741, | |
| "step": 3112 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.631897905617746e-07, | |
| "loss": 1.6569, | |
| "step": 3114 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.6210539057225214e-07, | |
| "loss": 1.6799, | |
| "step": 3116 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.610228921528234e-07, | |
| "loss": 1.6369, | |
| "step": 3118 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.599422980931093e-07, | |
| "loss": 1.6839, | |
| "step": 3120 | |
| }, | |
| { | |
| "epoch": 3.09, | |
| "learning_rate": 2.588636111778242e-07, | |
| "loss": 1.6174, | |
| "step": 3122 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.5778683418676683e-07, | |
| "loss": 1.6918, | |
| "step": 3124 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.5671196989481436e-07, | |
| "loss": 1.7307, | |
| "step": 3126 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.556390210719149e-07, | |
| "loss": 1.7038, | |
| "step": 3128 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.54567990483081e-07, | |
| "loss": 1.644, | |
| "step": 3130 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 2.5349888088838033e-07, | |
| "loss": 1.6354, | |
| "step": 3132 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.524316950429318e-07, | |
| "loss": 1.6278, | |
| "step": 3134 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.5136643569689587e-07, | |
| "loss": 1.7275, | |
| "step": 3136 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.5030310559546797e-07, | |
| "loss": 1.6689, | |
| "step": 3138 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.4924170747887275e-07, | |
| "loss": 1.726, | |
| "step": 3140 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 2.48182244082356e-07, | |
| "loss": 1.7026, | |
| "step": 3142 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.471247181361772e-07, | |
| "loss": 1.7054, | |
| "step": 3144 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.4606913236560277e-07, | |
| "loss": 1.6816, | |
| "step": 3146 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.450154894909001e-07, | |
| "loss": 1.6525, | |
| "step": 3148 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.439637922273294e-07, | |
| "loss": 1.7477, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 3.12, | |
| "learning_rate": 2.4291404328513686e-07, | |
| "loss": 1.6488, | |
| "step": 3152 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.4186624536954736e-07, | |
| "loss": 1.7179, | |
| "step": 3154 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.408204011807589e-07, | |
| "loss": 1.651, | |
| "step": 3156 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.3977651341393367e-07, | |
| "loss": 1.6154, | |
| "step": 3158 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.3873458475919296e-07, | |
| "loss": 1.6606, | |
| "step": 3160 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 2.376946179016094e-07, | |
| "loss": 1.6724, | |
| "step": 3162 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.366566155211992e-07, | |
| "loss": 1.7043, | |
| "step": 3164 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.3562058029291654e-07, | |
| "loss": 1.6914, | |
| "step": 3166 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.345865148866466e-07, | |
| "loss": 1.6487, | |
| "step": 3168 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.3355442196719732e-07, | |
| "loss": 1.6732, | |
| "step": 3170 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.3252430419429492e-07, | |
| "loss": 1.7267, | |
| "step": 3172 | |
| }, | |
| { | |
| "epoch": 3.14, | |
| "learning_rate": 2.314961642225739e-07, | |
| "loss": 1.6803, | |
| "step": 3174 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.3047000470157363e-07, | |
| "loss": 1.7008, | |
| "step": 3176 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.2944582827572835e-07, | |
| "loss": 1.6797, | |
| "step": 3178 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.2842363758436267e-07, | |
| "loss": 1.7207, | |
| "step": 3180 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.2740343526168448e-07, | |
| "loss": 1.7016, | |
| "step": 3182 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 2.2638522393677562e-07, | |
| "loss": 1.6971, | |
| "step": 3184 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.253690062335888e-07, | |
| "loss": 1.6489, | |
| "step": 3186 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.2435478477093917e-07, | |
| "loss": 1.6823, | |
| "step": 3188 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.2334256216249614e-07, | |
| "loss": 1.6691, | |
| "step": 3190 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.2233234101677967e-07, | |
| "loss": 1.6913, | |
| "step": 3192 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 2.213241239371505e-07, | |
| "loss": 1.5959, | |
| "step": 3194 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.2031791352180607e-07, | |
| "loss": 1.6952, | |
| "step": 3196 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.1931371236377128e-07, | |
| "loss": 1.6515, | |
| "step": 3198 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.1831152305089463e-07, | |
| "loss": 1.6253, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.1731134816583886e-07, | |
| "loss": 1.6846, | |
| "step": 3202 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 2.1631319028607565e-07, | |
| "loss": 1.6864, | |
| "step": 3204 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.153170519838795e-07, | |
| "loss": 1.6426, | |
| "step": 3206 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1432293582632012e-07, | |
| "loss": 1.6669, | |
| "step": 3208 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1333084437525562e-07, | |
| "loss": 1.7057, | |
| "step": 3210 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1234078018732671e-07, | |
| "loss": 1.7278, | |
| "step": 3212 | |
| }, | |
| { | |
| "epoch": 3.18, | |
| "learning_rate": 2.1135274581395025e-07, | |
| "loss": 1.6809, | |
| "step": 3214 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.1036674380131137e-07, | |
| "loss": 1.6837, | |
| "step": 3216 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0938277669035832e-07, | |
| "loss": 1.6744, | |
| "step": 3218 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0840084701679573e-07, | |
| "loss": 1.6669, | |
| "step": 3220 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.074209573110769e-07, | |
| "loss": 1.6984, | |
| "step": 3222 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 2.0644311009839798e-07, | |
| "loss": 1.6924, | |
| "step": 3224 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.054673078986926e-07, | |
| "loss": 1.7082, | |
| "step": 3226 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.0449355322662387e-07, | |
| "loss": 1.735, | |
| "step": 3228 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.0352184859157794e-07, | |
| "loss": 1.6493, | |
| "step": 3230 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.0255219649765832e-07, | |
| "loss": 1.6953, | |
| "step": 3232 | |
| }, | |
| { | |
| "epoch": 3.2, | |
| "learning_rate": 2.0158459944367924e-07, | |
| "loss": 1.7332, | |
| "step": 3234 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 2.0061905992315853e-07, | |
| "loss": 1.6583, | |
| "step": 3236 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.9965558042431264e-07, | |
| "loss": 1.6999, | |
| "step": 3238 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.9869416343004808e-07, | |
| "loss": 1.6864, | |
| "step": 3240 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.9773481141795755e-07, | |
| "loss": 1.6661, | |
| "step": 3242 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 1.967775268603109e-07, | |
| "loss": 1.6839, | |
| "step": 3244 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.9582231222405132e-07, | |
| "loss": 1.6569, | |
| "step": 3246 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.9486916997078706e-07, | |
| "loss": 1.6366, | |
| "step": 3248 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.9391810255678565e-07, | |
| "loss": 1.6826, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.9296911243296822e-07, | |
| "loss": 1.6749, | |
| "step": 3252 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 1.9202220204490293e-07, | |
| "loss": 1.7141, | |
| "step": 3254 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.9107737383279733e-07, | |
| "loss": 1.6661, | |
| "step": 3256 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.9013463023149435e-07, | |
| "loss": 1.6622, | |
| "step": 3258 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.8919397367046409e-07, | |
| "loss": 1.6893, | |
| "step": 3260 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.8825540657379812e-07, | |
| "loss": 1.6911, | |
| "step": 3262 | |
| }, | |
| { | |
| "epoch": 3.23, | |
| "learning_rate": 1.8731893136020415e-07, | |
| "loss": 1.6718, | |
| "step": 3264 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.8638455044299894e-07, | |
| "loss": 1.7332, | |
| "step": 3266 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.854522662301018e-07, | |
| "loss": 1.7031, | |
| "step": 3268 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.8452208112402857e-07, | |
| "loss": 1.647, | |
| "step": 3270 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.835939975218863e-07, | |
| "loss": 1.6536, | |
| "step": 3272 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 1.8266801781536655e-07, | |
| "loss": 1.6738, | |
| "step": 3274 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.8174414439073816e-07, | |
| "loss": 1.6868, | |
| "step": 3276 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.8082237962884306e-07, | |
| "loss": 1.6532, | |
| "step": 3278 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.7990272590508827e-07, | |
| "loss": 1.6979, | |
| "step": 3280 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.7898518558944088e-07, | |
| "loss": 1.6538, | |
| "step": 3282 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 1.7806976104642202e-07, | |
| "loss": 1.7394, | |
| "step": 3284 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.7715645463510041e-07, | |
| "loss": 1.6426, | |
| "step": 3286 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.7624526870908584e-07, | |
| "loss": 1.7107, | |
| "step": 3288 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.7533620561652363e-07, | |
| "loss": 1.6849, | |
| "step": 3290 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.7442926770008903e-07, | |
| "loss": 1.66, | |
| "step": 3292 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 1.7352445729697994e-07, | |
| "loss": 1.6332, | |
| "step": 3294 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.7262177673891253e-07, | |
| "loss": 1.7031, | |
| "step": 3296 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.7172122835211333e-07, | |
| "loss": 1.6407, | |
| "step": 3298 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.7082281445731495e-07, | |
| "loss": 1.705, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.6992653736974883e-07, | |
| "loss": 1.6543, | |
| "step": 3302 | |
| }, | |
| { | |
| "epoch": 3.27, | |
| "learning_rate": 1.6903239939914016e-07, | |
| "loss": 1.668, | |
| "step": 3304 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.681404028497021e-07, | |
| "loss": 1.6595, | |
| "step": 3306 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.6725055002012768e-07, | |
| "loss": 1.6746, | |
| "step": 3308 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.6636284320358707e-07, | |
| "loss": 1.7035, | |
| "step": 3310 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.6547728468771983e-07, | |
| "loss": 1.729, | |
| "step": 3312 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 1.6459387675462867e-07, | |
| "loss": 1.6552, | |
| "step": 3314 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.6371262168087497e-07, | |
| "loss": 1.6789, | |
| "step": 3316 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.6283352173747146e-07, | |
| "loss": 1.6701, | |
| "step": 3318 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.6195657918987758e-07, | |
| "loss": 1.5709, | |
| "step": 3320 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.610817962979928e-07, | |
| "loss": 1.6626, | |
| "step": 3322 | |
| }, | |
| { | |
| "epoch": 3.29, | |
| "learning_rate": 1.6064521542763833e-07, | |
| "loss": 1.6294, | |
| "step": 3324 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.5977367624445305e-07, | |
| "loss": 1.6654, | |
| "step": 3326 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.589043023423643e-07, | |
| "loss": 1.6842, | |
| "step": 3328 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.5847042807207256e-07, | |
| "loss": 1.7304, | |
| "step": 3330 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.576043062906236e-07, | |
| "loss": 1.6603, | |
| "step": 3332 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 1.567403553807838e-07, | |
| "loss": 1.6465, | |
| "step": 3334 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.5587857756897283e-07, | |
| "loss": 1.6381, | |
| "step": 3336 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.5501897507601015e-07, | |
| "loss": 1.7105, | |
| "step": 3338 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.5416155011711018e-07, | |
| "loss": 1.7057, | |
| "step": 3340 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.5330630490187468e-07, | |
| "loss": 1.6832, | |
| "step": 3342 | |
| }, | |
| { | |
| "epoch": 3.31, | |
| "learning_rate": 1.5245324163428908e-07, | |
| "loss": 1.6965, | |
| "step": 3344 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.5160236251271574e-07, | |
| "loss": 1.683, | |
| "step": 3346 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.5075366972988812e-07, | |
| "loss": 1.677, | |
| "step": 3348 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.4990716547290517e-07, | |
| "loss": 1.6961, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.4906285192322666e-07, | |
| "loss": 1.6437, | |
| "step": 3352 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 1.4822073125666678e-07, | |
| "loss": 1.6676, | |
| "step": 3354 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4738080564338817e-07, | |
| "loss": 1.7125, | |
| "step": 3356 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4654307724789682e-07, | |
| "loss": 1.6495, | |
| "step": 3358 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4570754822903686e-07, | |
| "loss": 1.7064, | |
| "step": 3360 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.448742207399841e-07, | |
| "loss": 1.6487, | |
| "step": 3362 | |
| }, | |
| { | |
| "epoch": 3.33, | |
| "learning_rate": 1.4404309692824135e-07, | |
| "loss": 1.6539, | |
| "step": 3364 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.4321417893563258e-07, | |
| "loss": 1.6744, | |
| "step": 3366 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.4238746889829678e-07, | |
| "loss": 1.644, | |
| "step": 3368 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.4156296894668317e-07, | |
| "loss": 1.718, | |
| "step": 3370 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.4074068120554595e-07, | |
| "loss": 1.6806, | |
| "step": 3372 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 1.3992060779393778e-07, | |
| "loss": 1.7006, | |
| "step": 3374 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.3910275082520572e-07, | |
| "loss": 1.6705, | |
| "step": 3376 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.38287112406984e-07, | |
| "loss": 1.6357, | |
| "step": 3378 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.3747369464119074e-07, | |
| "loss": 1.6412, | |
| "step": 3380 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.366624996240203e-07, | |
| "loss": 1.7191, | |
| "step": 3382 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.358535294459401e-07, | |
| "loss": 1.7351, | |
| "step": 3384 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 1.3504678619168297e-07, | |
| "loss": 1.6871, | |
| "step": 3386 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3424227194024328e-07, | |
| "loss": 1.6774, | |
| "step": 3388 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3343998876487172e-07, | |
| "loss": 1.6983, | |
| "step": 3390 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3263993873306923e-07, | |
| "loss": 1.6706, | |
| "step": 3392 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.3184212390658155e-07, | |
| "loss": 1.7114, | |
| "step": 3394 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 1.310465463413941e-07, | |
| "loss": 1.6966, | |
| "step": 3396 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.3025320808772766e-07, | |
| "loss": 1.697, | |
| "step": 3398 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2946211119003169e-07, | |
| "loss": 1.6294, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2867325768697946e-07, | |
| "loss": 1.7066, | |
| "step": 3402 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2788664961146367e-07, | |
| "loss": 1.6652, | |
| "step": 3404 | |
| }, | |
| { | |
| "epoch": 3.37, | |
| "learning_rate": 1.2710228899058971e-07, | |
| "loss": 1.6734, | |
| "step": 3406 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.2632017784567138e-07, | |
| "loss": 1.724, | |
| "step": 3408 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.25540318192226e-07, | |
| "loss": 1.6693, | |
| "step": 3410 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.2476271203996868e-07, | |
| "loss": 1.676, | |
| "step": 3412 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.2398736139280687e-07, | |
| "loss": 1.719, | |
| "step": 3414 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 1.2321426824883517e-07, | |
| "loss": 1.7009, | |
| "step": 3416 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.224434346003318e-07, | |
| "loss": 1.6509, | |
| "step": 3418 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.2167486243375092e-07, | |
| "loss": 1.6845, | |
| "step": 3420 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.2090855372971942e-07, | |
| "loss": 1.6586, | |
| "step": 3422 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.2014451046303142e-07, | |
| "loss": 1.6796, | |
| "step": 3424 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 1.193827346026426e-07, | |
| "loss": 1.6421, | |
| "step": 3426 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1862322811166514e-07, | |
| "loss": 1.6503, | |
| "step": 3428 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1786599294736366e-07, | |
| "loss": 1.716, | |
| "step": 3430 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1711103106114961e-07, | |
| "loss": 1.696, | |
| "step": 3432 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.1635834439857528e-07, | |
| "loss": 1.6932, | |
| "step": 3434 | |
| }, | |
| { | |
| "epoch": 3.4, | |
| "learning_rate": 1.156079348993303e-07, | |
| "loss": 1.6951, | |
| "step": 3436 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.148598044972363e-07, | |
| "loss": 1.6551, | |
| "step": 3438 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.1411395512024069e-07, | |
| "loss": 1.7018, | |
| "step": 3440 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.133703886904136e-07, | |
| "loss": 1.6772, | |
| "step": 3442 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.1262910712394125e-07, | |
| "loss": 1.6992, | |
| "step": 3444 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 1.1189011233112223e-07, | |
| "loss": 1.6705, | |
| "step": 3446 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.111534062163616e-07, | |
| "loss": 1.6787, | |
| "step": 3448 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.1041899067816663e-07, | |
| "loss": 1.7002, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.0968686760914248e-07, | |
| "loss": 1.6665, | |
| "step": 3452 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.0895703889598473e-07, | |
| "loss": 1.6806, | |
| "step": 3454 | |
| }, | |
| { | |
| "epoch": 3.42, | |
| "learning_rate": 1.0822950641947814e-07, | |
| "loss": 1.6748, | |
| "step": 3456 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0750427205448953e-07, | |
| "loss": 1.6429, | |
| "step": 3458 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0678133766996277e-07, | |
| "loss": 1.6581, | |
| "step": 3460 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0606070512891562e-07, | |
| "loss": 1.6574, | |
| "step": 3462 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0534237628843323e-07, | |
| "loss": 1.6619, | |
| "step": 3464 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 1.0462635299966404e-07, | |
| "loss": 1.6999, | |
| "step": 3466 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.0391263710781561e-07, | |
| "loss": 1.6289, | |
| "step": 3468 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.0320123045214902e-07, | |
| "loss": 1.6939, | |
| "step": 3470 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.024921348659743e-07, | |
| "loss": 1.7441, | |
| "step": 3472 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.0178535217664552e-07, | |
| "loss": 1.697, | |
| "step": 3474 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 1.0108088420555694e-07, | |
| "loss": 1.6491, | |
| "step": 3476 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 1.0037873276813769e-07, | |
| "loss": 1.6462, | |
| "step": 3478 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.967889967384668e-08, | |
| "loss": 1.6783, | |
| "step": 3480 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.898138672616863e-08, | |
| "loss": 1.691, | |
| "step": 3482 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.828619572260954e-08, | |
| "loss": 1.723, | |
| "step": 3484 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 9.759332845469104e-08, | |
| "loss": 1.683, | |
| "step": 3486 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.690278670794682e-08, | |
| "loss": 1.6886, | |
| "step": 3488 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.62145722619182e-08, | |
| "loss": 1.7261, | |
| "step": 3490 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.552868689014792e-08, | |
| "loss": 1.6743, | |
| "step": 3492 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.48451323601771e-08, | |
| "loss": 1.6576, | |
| "step": 3494 | |
| }, | |
| { | |
| "epoch": 3.46, | |
| "learning_rate": 9.416391043354043e-08, | |
| "loss": 1.6683, | |
| "step": 3496 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.348502286576154e-08, | |
| "loss": 1.6587, | |
| "step": 3498 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.280847140634796e-08, | |
| "loss": 1.683, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.213425779878692e-08, | |
| "loss": 1.6838, | |
| "step": 3502 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.14623837805416e-08, | |
| "loss": 1.6175, | |
| "step": 3504 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 9.079285108304535e-08, | |
| "loss": 1.6349, | |
| "step": 3506 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 9.01256614316982e-08, | |
| "loss": 1.6756, | |
| "step": 3508 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.946081654586258e-08, | |
| "loss": 1.6785, | |
| "step": 3510 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.879831813885697e-08, | |
| "loss": 1.6431, | |
| "step": 3512 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.813816791795448e-08, | |
| "loss": 1.6858, | |
| "step": 3514 | |
| }, | |
| { | |
| "epoch": 3.48, | |
| "learning_rate": 8.748036758437616e-08, | |
| "loss": 1.6935, | |
| "step": 3516 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.682491883328724e-08, | |
| "loss": 1.6505, | |
| "step": 3518 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.617182335379346e-08, | |
| "loss": 1.6802, | |
| "step": 3520 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.552108282893545e-08, | |
| "loss": 1.695, | |
| "step": 3522 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.487269893568549e-08, | |
| "loss": 1.6742, | |
| "step": 3524 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 8.422667334494249e-08, | |
| "loss": 1.7183, | |
| "step": 3526 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.358300772152849e-08, | |
| "loss": 1.6844, | |
| "step": 3528 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.29417037241833e-08, | |
| "loss": 1.6702, | |
| "step": 3530 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.230276300556071e-08, | |
| "loss": 1.6505, | |
| "step": 3532 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.166618721222462e-08, | |
| "loss": 1.6815, | |
| "step": 3534 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 8.103197798464478e-08, | |
| "loss": 1.6861, | |
| "step": 3536 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 8.040013695719162e-08, | |
| "loss": 1.6608, | |
| "step": 3538 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.977066575813263e-08, | |
| "loss": 1.6985, | |
| "step": 3540 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.914356600962879e-08, | |
| "loss": 1.7021, | |
| "step": 3542 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.851883932772929e-08, | |
| "loss": 1.6672, | |
| "step": 3544 | |
| }, | |
| { | |
| "epoch": 3.51, | |
| "learning_rate": 7.78964873223682e-08, | |
| "loss": 1.6433, | |
| "step": 3546 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.727651159735992e-08, | |
| "loss": 1.6498, | |
| "step": 3548 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.665891375039501e-08, | |
| "loss": 1.6734, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.6043695373036e-08, | |
| "loss": 1.6656, | |
| "step": 3552 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.543085805071392e-08, | |
| "loss": 1.6649, | |
| "step": 3554 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 7.482040336272388e-08, | |
| "loss": 1.6739, | |
| "step": 3556 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.421233288222029e-08, | |
| "loss": 1.7235, | |
| "step": 3558 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.360664817621342e-08, | |
| "loss": 1.6998, | |
| "step": 3560 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.30033508055663e-08, | |
| "loss": 1.7018, | |
| "step": 3562 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.240244232498837e-08, | |
| "loss": 1.7032, | |
| "step": 3564 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 7.180392428303394e-08, | |
| "loss": 1.6909, | |
| "step": 3566 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 7.120779822209688e-08, | |
| "loss": 1.7179, | |
| "step": 3568 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 7.061406567840655e-08, | |
| "loss": 1.6764, | |
| "step": 3570 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 7.002272818202415e-08, | |
| "loss": 1.7069, | |
| "step": 3572 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 6.943378725683912e-08, | |
| "loss": 1.7024, | |
| "step": 3574 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 6.884724442056499e-08, | |
| "loss": 1.6531, | |
| "step": 3576 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.826310118473466e-08, | |
| "loss": 1.6295, | |
| "step": 3578 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.768135905469785e-08, | |
| "loss": 1.6435, | |
| "step": 3580 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.71020195296167e-08, | |
| "loss": 1.6703, | |
| "step": 3582 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.652508410246094e-08, | |
| "loss": 1.6818, | |
| "step": 3584 | |
| }, | |
| { | |
| "epoch": 3.55, | |
| "learning_rate": 6.595055426000595e-08, | |
| "loss": 1.68, | |
| "step": 3586 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.537843148282706e-08, | |
| "loss": 1.6795, | |
| "step": 3588 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.480871724529668e-08, | |
| "loss": 1.7166, | |
| "step": 3590 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.424141301558084e-08, | |
| "loss": 1.6654, | |
| "step": 3592 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.367652025563453e-08, | |
| "loss": 1.7272, | |
| "step": 3594 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 6.311404042119828e-08, | |
| "loss": 1.6226, | |
| "step": 3596 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.255397496179448e-08, | |
| "loss": 1.7252, | |
| "step": 3598 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.199632532072397e-08, | |
| "loss": 1.7283, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.144109293506172e-08, | |
| "loss": 1.6501, | |
| "step": 3602 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.088827923565321e-08, | |
| "loss": 1.7162, | |
| "step": 3604 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 6.033788564711128e-08, | |
| "loss": 1.6611, | |
| "step": 3606 | |
| }, | |
| { | |
| "epoch": 3.57, | |
| "learning_rate": 5.978991358781182e-08, | |
| "loss": 1.6925, | |
| "step": 3608 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.924436446989034e-08, | |
| "loss": 1.6528, | |
| "step": 3610 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.870123969923868e-08, | |
| "loss": 1.7173, | |
| "step": 3612 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.8160540675500913e-08, | |
| "loss": 1.6828, | |
| "step": 3614 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.7622268792069996e-08, | |
| "loss": 1.6779, | |
| "step": 3616 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 5.7086425436083775e-08, | |
| "loss": 1.6229, | |
| "step": 3618 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.6553011988421993e-08, | |
| "loss": 1.6758, | |
| "step": 3620 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.6022029823702745e-08, | |
| "loss": 1.667, | |
| "step": 3622 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.5493480310278115e-08, | |
| "loss": 1.6687, | |
| "step": 3624 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.496736481023145e-08, | |
| "loss": 1.7242, | |
| "step": 3626 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 5.444368467937366e-08, | |
| "loss": 1.6859, | |
| "step": 3628 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.392244126723933e-08, | |
| "loss": 1.6798, | |
| "step": 3630 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.340363591708408e-08, | |
| "loss": 1.6622, | |
| "step": 3632 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.2887269965880556e-08, | |
| "loss": 1.6727, | |
| "step": 3634 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.2373344744314095e-08, | |
| "loss": 1.7243, | |
| "step": 3636 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 5.1861861576781405e-08, | |
| "loss": 1.7299, | |
| "step": 3638 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 5.135282178138545e-08, | |
| "loss": 1.6512, | |
| "step": 3640 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 5.0846226669932437e-08, | |
| "loss": 1.641, | |
| "step": 3642 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 5.034207754792896e-08, | |
| "loss": 1.664, | |
| "step": 3644 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 4.9840375714577666e-08, | |
| "loss": 1.6698, | |
| "step": 3646 | |
| }, | |
| { | |
| "epoch": 3.61, | |
| "learning_rate": 4.934112246277533e-08, | |
| "loss": 1.7022, | |
| "step": 3648 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.88443190791078e-08, | |
| "loss": 1.6603, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.834996684384795e-08, | |
| "loss": 1.7094, | |
| "step": 3652 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.785806703095252e-08, | |
| "loss": 1.6443, | |
| "step": 3654 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.7368620908057155e-08, | |
| "loss": 1.6811, | |
| "step": 3656 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 4.6881629736475137e-08, | |
| "loss": 1.6409, | |
| "step": 3658 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.639709477119325e-08, | |
| "loss": 1.6218, | |
| "step": 3660 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.59150172608681e-08, | |
| "loss": 1.701, | |
| "step": 3662 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.543539844782396e-08, | |
| "loss": 1.669, | |
| "step": 3664 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.495823956804834e-08, | |
| "loss": 1.704, | |
| "step": 3666 | |
| }, | |
| { | |
| "epoch": 3.63, | |
| "learning_rate": 4.4483541851190297e-08, | |
| "loss": 1.649, | |
| "step": 3668 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.4011306520555514e-08, | |
| "loss": 1.6022, | |
| "step": 3670 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.354153479310485e-08, | |
| "loss": 1.6646, | |
| "step": 3672 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.30742278794497e-08, | |
| "loss": 1.6509, | |
| "step": 3674 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.260938698384997e-08, | |
| "loss": 1.6517, | |
| "step": 3676 | |
| }, | |
| { | |
| "epoch": 3.64, | |
| "learning_rate": 4.2147013304210666e-08, | |
| "loss": 1.6678, | |
| "step": 3678 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.168710803207864e-08, | |
| "loss": 1.701, | |
| "step": 3680 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.122967235263941e-08, | |
| "loss": 1.689, | |
| "step": 3682 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.077470744471434e-08, | |
| "loss": 1.6651, | |
| "step": 3684 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 4.03222144807579e-08, | |
| "loss": 1.6899, | |
| "step": 3686 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 3.987219462685387e-08, | |
| "loss": 1.7042, | |
| "step": 3688 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.942464904271314e-08, | |
| "loss": 1.6799, | |
| "step": 3690 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.897957888167025e-08, | |
| "loss": 1.6515, | |
| "step": 3692 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.853698529068028e-08, | |
| "loss": 1.7065, | |
| "step": 3694 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.8096869410316426e-08, | |
| "loss": 1.6632, | |
| "step": 3696 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 3.765923237476654e-08, | |
| "loss": 1.6718, | |
| "step": 3698 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.72240753118308e-08, | |
| "loss": 1.6799, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.6791399342918086e-08, | |
| "loss": 1.7134, | |
| "step": 3702 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.6361205583043456e-08, | |
| "loss": 1.6646, | |
| "step": 3704 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.593349514082555e-08, | |
| "loss": 1.6454, | |
| "step": 3706 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 3.5508269118483034e-08, | |
| "loss": 1.7319, | |
| "step": 3708 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.508552861183267e-08, | |
| "loss": 1.6319, | |
| "step": 3710 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.46652747102858e-08, | |
| "loss": 1.7415, | |
| "step": 3712 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.4247508496845326e-08, | |
| "loss": 1.6909, | |
| "step": 3714 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.3832231048103735e-08, | |
| "loss": 1.6501, | |
| "step": 3716 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 3.341944343424008e-08, | |
| "loss": 1.7375, | |
| "step": 3718 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.300914671901656e-08, | |
| "loss": 1.6772, | |
| "step": 3720 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.2601341959776594e-08, | |
| "loss": 1.6978, | |
| "step": 3722 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.219603020744177e-08, | |
| "loss": 1.649, | |
| "step": 3724 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.179321250650924e-08, | |
| "loss": 1.6255, | |
| "step": 3726 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 3.13928898950484e-08, | |
| "loss": 1.6701, | |
| "step": 3728 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 3.099506340469948e-08, | |
| "loss": 1.7098, | |
| "step": 3730 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 3.0599734060669626e-08, | |
| "loss": 1.6822, | |
| "step": 3732 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 3.0206902881731e-08, | |
| "loss": 1.7008, | |
| "step": 3734 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 2.981657088021783e-08, | |
| "loss": 1.6697, | |
| "step": 3736 | |
| }, | |
| { | |
| "epoch": 3.7, | |
| "learning_rate": 2.9428739062024144e-08, | |
| "loss": 1.6766, | |
| "step": 3738 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.904340842660069e-08, | |
| "loss": 1.6991, | |
| "step": 3740 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.86605799669527e-08, | |
| "loss": 1.6161, | |
| "step": 3742 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.8280254669637127e-08, | |
| "loss": 1.6393, | |
| "step": 3744 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.7902433514760626e-08, | |
| "loss": 1.6797, | |
| "step": 3746 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 2.7527117475976136e-08, | |
| "loss": 1.6685, | |
| "step": 3748 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.71543075204812e-08, | |
| "loss": 1.6945, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.6784004609014864e-08, | |
| "loss": 1.6821, | |
| "step": 3752 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.6416209695855563e-08, | |
| "loss": 1.7373, | |
| "step": 3754 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.6050923728818784e-08, | |
| "loss": 1.6551, | |
| "step": 3756 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 2.56881476492542e-08, | |
| "loss": 1.6936, | |
| "step": 3758 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.532788239204342e-08, | |
| "loss": 1.7145, | |
| "step": 3760 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.4970128885597576e-08, | |
| "loss": 1.7068, | |
| "step": 3762 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.4614888051855297e-08, | |
| "loss": 1.7091, | |
| "step": 3764 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.426216080627952e-08, | |
| "loss": 1.6595, | |
| "step": 3766 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 2.3911948057856125e-08, | |
| "loss": 1.6309, | |
| "step": 3768 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.3564250709090405e-08, | |
| "loss": 1.6751, | |
| "step": 3770 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.3219069656006285e-08, | |
| "loss": 1.7125, | |
| "step": 3772 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.2876405788142207e-08, | |
| "loss": 1.6595, | |
| "step": 3774 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.2536259988550577e-08, | |
| "loss": 1.6482, | |
| "step": 3776 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 2.2198633133794444e-08, | |
| "loss": 1.6452, | |
| "step": 3778 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.186352609394504e-08, | |
| "loss": 1.6772, | |
| "step": 3780 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.1530939732580466e-08, | |
| "loss": 1.692, | |
| "step": 3782 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.1200874906782905e-08, | |
| "loss": 1.6542, | |
| "step": 3784 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.0873332467136405e-08, | |
| "loss": 1.692, | |
| "step": 3786 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 2.0548313257724882e-08, | |
| "loss": 1.6492, | |
| "step": 3788 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 2.0225818116129445e-08, | |
| "loss": 1.6871, | |
| "step": 3790 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.9905847873427305e-08, | |
| "loss": 1.6905, | |
| "step": 3792 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.958840335418832e-08, | |
| "loss": 1.6602, | |
| "step": 3794 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.9273485376473997e-08, | |
| "loss": 1.6722, | |
| "step": 3796 | |
| }, | |
| { | |
| "epoch": 3.76, | |
| "learning_rate": 1.8961094751834382e-08, | |
| "loss": 1.6703, | |
| "step": 3798 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.8651232285306962e-08, | |
| "loss": 1.666, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.8343898775413647e-08, | |
| "loss": 1.6542, | |
| "step": 3802 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.8039095014159788e-08, | |
| "loss": 1.6748, | |
| "step": 3804 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.7736821787031063e-08, | |
| "loss": 1.6764, | |
| "step": 3806 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 1.74370798729917e-08, | |
| "loss": 1.681, | |
| "step": 3808 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.713987004448325e-08, | |
| "loss": 1.6905, | |
| "step": 3810 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.6845193067421713e-08, | |
| "loss": 1.661, | |
| "step": 3812 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.6553049701195865e-08, | |
| "loss": 1.6492, | |
| "step": 3814 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.6263440698665587e-08, | |
| "loss": 1.7071, | |
| "step": 3816 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.5976366806159215e-08, | |
| "loss": 1.7526, | |
| "step": 3818 | |
| }, | |
| { | |
| "epoch": 3.78, | |
| "learning_rate": 1.569182876347219e-08, | |
| "loss": 1.6962, | |
| "step": 3820 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.540982730386531e-08, | |
| "loss": 1.689, | |
| "step": 3822 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.5130363154062244e-08, | |
| "loss": 1.6757, | |
| "step": 3824 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.4853437034248129e-08, | |
| "loss": 1.7157, | |
| "step": 3826 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.4579049658067333e-08, | |
| "loss": 1.7109, | |
| "step": 3828 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 1.430720173262201e-08, | |
| "loss": 1.6854, | |
| "step": 3830 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.4037893958469993e-08, | |
| "loss": 1.617, | |
| "step": 3832 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.3771127029623132e-08, | |
| "loss": 1.6639, | |
| "step": 3834 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.350690163354573e-08, | |
| "loss": 1.6558, | |
| "step": 3836 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.3245218451152008e-08, | |
| "loss": 1.649, | |
| "step": 3838 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 1.2986078156804969e-08, | |
| "loss": 1.6518, | |
| "step": 3840 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.272948141831498e-08, | |
| "loss": 1.6551, | |
| "step": 3842 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.247542889693709e-08, | |
| "loss": 1.6744, | |
| "step": 3844 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.2223921247370151e-08, | |
| "loss": 1.6261, | |
| "step": 3846 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.1974959117754591e-08, | |
| "loss": 1.7008, | |
| "step": 3848 | |
| }, | |
| { | |
| "epoch": 3.81, | |
| "learning_rate": 1.17285431496712e-08, | |
| "loss": 1.6909, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.1484673978139125e-08, | |
| "loss": 1.6798, | |
| "step": 3852 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.1243352231614434e-08, | |
| "loss": 1.6802, | |
| "step": 3854 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.100457853198844e-08, | |
| "loss": 1.6862, | |
| "step": 3856 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.0768353494585935e-08, | |
| "loss": 1.6948, | |
| "step": 3858 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 1.0534677728163855e-08, | |
| "loss": 1.7023, | |
| "step": 3860 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 1.030355183490983e-08, | |
| "loss": 1.7078, | |
| "step": 3862 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 1.0074976410439972e-08, | |
| "loss": 1.6903, | |
| "step": 3864 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 9.848952043798097e-09, | |
| "loss": 1.6432, | |
| "step": 3866 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 9.625479317453722e-09, | |
| "loss": 1.6835, | |
| "step": 3868 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 9.404558807301065e-09, | |
| "loss": 1.6456, | |
| "step": 3870 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 9.186191082656613e-09, | |
| "loss": 1.6611, | |
| "step": 3872 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.970376706258886e-09, | |
| "loss": 1.6522, | |
| "step": 3874 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.757116234266115e-09, | |
| "loss": 1.6628, | |
| "step": 3876 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.546410216254907e-09, | |
| "loss": 1.6732, | |
| "step": 3878 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 8.338259195219243e-09, | |
| "loss": 1.681, | |
| "step": 3880 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 8.132663707568821e-09, | |
| "loss": 1.6631, | |
| "step": 3882 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 7.929624283127489e-09, | |
| "loss": 1.6971, | |
| "step": 3884 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 7.729141445132038e-09, | |
| "loss": 1.6769, | |
| "step": 3886 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 7.531215710231298e-09, | |
| "loss": 1.6463, | |
| "step": 3888 | |
| }, | |
| { | |
| "epoch": 3.85, | |
| "learning_rate": 7.335847588483823e-09, | |
| "loss": 1.6906, | |
| "step": 3890 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 7.1430375833574365e-09, | |
| "loss": 1.6887, | |
| "step": 3892 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 6.952786191727899e-09, | |
| "loss": 1.6517, | |
| "step": 3894 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 6.765093903876806e-09, | |
| "loss": 1.7011, | |
| "step": 3896 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 6.5799612034912465e-09, | |
| "loss": 1.6609, | |
| "step": 3898 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 6.397388567662143e-09, | |
| "loss": 1.7013, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 6.21737646688314e-09, | |
| "loss": 1.6742, | |
| "step": 3902 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 6.0399253650490474e-09, | |
| "loss": 1.6729, | |
| "step": 3904 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 5.86503571945518e-09, | |
| "loss": 1.7014, | |
| "step": 3906 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 5.69270798079613e-09, | |
| "loss": 1.6966, | |
| "step": 3908 | |
| }, | |
| { | |
| "epoch": 3.87, | |
| "learning_rate": 5.522942593163771e-09, | |
| "loss": 1.6772, | |
| "step": 3910 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 5.355739994047481e-09, | |
| "loss": 1.6908, | |
| "step": 3912 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 5.19110061433159e-09, | |
| "loss": 1.6486, | |
| "step": 3914 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 5.029024878295373e-09, | |
| "loss": 1.6703, | |
| "step": 3916 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.86951320361162e-09, | |
| "loss": 1.6672, | |
| "step": 3918 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 4.712566001345064e-09, | |
| "loss": 1.6561, | |
| "step": 3920 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.558183675952176e-09, | |
| "loss": 1.6885, | |
| "step": 3922 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.406366625279378e-09, | |
| "loss": 1.7056, | |
| "step": 3924 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.257115240562381e-09, | |
| "loss": 1.6483, | |
| "step": 3926 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 4.110429906425406e-09, | |
| "loss": 1.7079, | |
| "step": 3928 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 3.966311000879408e-09, | |
| "loss": 1.6728, | |
| "step": 3930 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.824758895322078e-09, | |
| "loss": 1.6286, | |
| "step": 3932 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.6857739545360643e-09, | |
| "loss": 1.6412, | |
| "step": 3934 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.54935653668853e-09, | |
| "loss": 1.6989, | |
| "step": 3936 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.415506993330153e-09, | |
| "loss": 1.6291, | |
| "step": 3938 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 3.284225669394014e-09, | |
| "loss": 1.6346, | |
| "step": 3940 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 3.1555129031948235e-09, | |
| "loss": 1.7289, | |
| "step": 3942 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 3.0293690264281413e-09, | |
| "loss": 1.6947, | |
| "step": 3944 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 2.9057943641693784e-09, | |
| "loss": 1.6906, | |
| "step": 3946 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 2.7847892348731306e-09, | |
| "loss": 1.6827, | |
| "step": 3948 | |
| }, | |
| { | |
| "epoch": 3.91, | |
| "learning_rate": 2.666353950372291e-09, | |
| "loss": 1.712, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.5504888158768277e-09, | |
| "loss": 1.6933, | |
| "step": 3952 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.437194129973896e-09, | |
| "loss": 1.6748, | |
| "step": 3954 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.3264701846261727e-09, | |
| "loss": 1.6532, | |
| "step": 3956 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.218317265171743e-09, | |
| "loss": 1.7141, | |
| "step": 3958 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 2.112735650322883e-09, | |
| "loss": 1.68, | |
| "step": 3960 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 2.0097256121657246e-09, | |
| "loss": 1.7106, | |
| "step": 3962 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.9092874161591442e-09, | |
| "loss": 1.7274, | |
| "step": 3964 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.811421321134765e-09, | |
| "loss": 1.6989, | |
| "step": 3966 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.716127579295401e-09, | |
| "loss": 1.714, | |
| "step": 3968 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 1.62340643621528e-09, | |
| "loss": 1.668, | |
| "step": 3970 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.5332581308384885e-09, | |
| "loss": 1.7128, | |
| "step": 3972 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.4456828954791945e-09, | |
| "loss": 1.6977, | |
| "step": 3974 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.360680955820759e-09, | |
| "loss": 1.7203, | |
| "step": 3976 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.2782525309149582e-09, | |
| "loss": 1.685, | |
| "step": 3978 | |
| }, | |
| { | |
| "epoch": 3.94, | |
| "learning_rate": 1.1983978331815413e-09, | |
| "loss": 1.6471, | |
| "step": 3980 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 1.1211170684076731e-09, | |
| "loss": 1.688, | |
| "step": 3982 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 1.0464104357477132e-09, | |
| "loss": 1.6823, | |
| "step": 3984 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 9.742781277221057e-10, | |
| "loss": 1.7105, | |
| "step": 3986 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 9.0472033021749e-10, | |
| "loss": 1.6584, | |
| "step": 3988 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 8.377372224855905e-10, | |
| "loss": 1.7026, | |
| "step": 3990 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 7.733289771434392e-10, | |
| "loss": 1.6724, | |
| "step": 3992 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 7.114957601723759e-10, | |
| "loss": 1.6822, | |
| "step": 3994 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 6.522377309180482e-10, | |
| "loss": 1.6495, | |
| "step": 3996 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 5.955550420895239e-10, | |
| "loss": 1.6512, | |
| "step": 3998 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 5.414478397592904e-10, | |
| "loss": 1.6595, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 4.899162633629217e-10, | |
| "loss": 1.6704, | |
| "step": 4002 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 4.409604456981908e-10, | |
| "loss": 1.6712, | |
| "step": 4004 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 3.9458051292529104e-10, | |
| "loss": 1.63, | |
| "step": 4006 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 3.507765845665034e-10, | |
| "loss": 1.7009, | |
| "step": 4008 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 3.0954877350519716e-10, | |
| "loss": 1.6726, | |
| "step": 4010 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 2.7089718598660715e-10, | |
| "loss": 1.6663, | |
| "step": 4012 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 2.348219216163905e-10, | |
| "loss": 1.6915, | |
| "step": 4014 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 2.0132307336151455e-10, | |
| "loss": 1.6691, | |
| "step": 4016 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 1.7040072754914702e-10, | |
| "loss": 1.6769, | |
| "step": 4018 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 1.4205496386687775e-10, | |
| "loss": 1.6206, | |
| "step": 4020 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 1.1628585536216374e-10, | |
| "loss": 1.6741, | |
| "step": 4022 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 9.30934684427731e-11, | |
| "loss": 1.6694, | |
| "step": 4024 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 7.247786287578605e-11, | |
| "loss": 1.6933, | |
| "step": 4026 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 5.443909178826089e-11, | |
| "loss": 1.685, | |
| "step": 4028 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 3.897720166623486e-11, | |
| "loss": 1.6814, | |
| "step": 4030 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 2.609223235561231e-11, | |
| "loss": 1.707, | |
| "step": 4032 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 1.5784217061054483e-11, | |
| "loss": 1.7169, | |
| "step": 4034 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "learning_rate": 8.053182346534626e-12, | |
| "loss": 1.6614, | |
| "step": 4036 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "step": 4036, | |
| "total_flos": 2.3255819414300262e+17, | |
| "train_loss": 1.7244027754360196, | |
| "train_runtime": 42261.7031, | |
| "train_samples_per_second": 6.113, | |
| "train_steps_per_second": 0.096 | |
| } | |
| ], | |
| "logging_steps": 2, | |
| "max_steps": 4036, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 4, | |
| "save_steps": 4000, | |
| "total_flos": 2.3255819414300262e+17, | |
| "train_batch_size": 4, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |